Non-Nuclear Methods for Compaction Control of Unbound Materials (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

47 chapter four METHODS FOR MEASURING THE IN SITU STIFFNESS/STRENGTH OF UNBOUND MATERIALS INTRODUCTION Although the density measurement has been long used for compaction control, it does not reflect the engineering prop- erties of unbound materials necessary to ensure their optimal performance. The key functional properties of soil layers are their stiffness and strength, which are considered to be mea- sures of their stability and resistance to deformation under load. Although the stiffness of a material defines its resistance to deformation before failure, the strength is its limiting stress value at failure. Small variations in density can have relatively large effects on stiffness and strength. Therefore, the errors that accumulate during the specified density procedure have the potential to significantly influence the performance of compacted unbound materials (White et al. 2007a). Stiffness and strength are also sensitive to variations in the moisture content, degree of saturation, and state of stress of compacted unbound materials, which all govern the mechanical behavior and response of these materials. In recent years, the shift from empirical to mechanistic– empirical pavement design procedures has resulted in a grow- ing interest in moving toward compaction control specifica- tions that emphasize stiffness and strength. This has led to the development of several in situ test devices that can measure the stiffness or strength of compacted unbound materials. These devices can be divided into four main groups. The first group consists of impact devices, such as the dynamic cone penetrometer (DCP) and the Clegg hammer (CH). The sec- ond category consists of devices that apply static, vibratory, or impact load to the ground, then estimate the stiffness based on the load and displacement measurements (using veloc- ity transducers or accelerometers); these devices include the Briaud compaction device (BCD), the GeoGauge, and the light weight deflectometer (LWD). A third group includes devices that are based on geophysical techniques and includes the portable seismic property analyzer (PSPA), in which sur- face waves are generated and detected in the tested layer to determine its modulus. Finally, the fourth group consists of sacrificial sensors that are buried in the compacted soil to monitor the growth in amplitude of compression waves dur- ing compaction. In addition to the previous in situ spot tests, technologies that provide continuous assessment of compaction, such as continuous compaction control (CCC) and intelligent com- paction (IC), have been investigated by DOTs as a viable tool for controlling the quality of compaction of various pave- ment layers and subgrade soils. This chapter summarizes information collected through lit- erature review of the performance of various in situ test devices and methods that have been evaluated by state DOTs to mea- sure stiffness, strength, or any parameter other than density for use in compaction control of unbound materials. For each device, the principle of operation, influence depth, reliabil- ity of measurement, and advantages and limitations are first provided. In addition, a summary of the main findings of pre- viously conducted studies is presented. In the preceding sec- tions, pictures of in situ devices from certain manufacturers are provided for demonstration purposes only. Inclusion of photos of these devices should not be construed as endorse- ments of the devices by this synthesis study. CLEGG HAMMER The CH was developed in Australia in late 1960s to measure the stiffness/strength of soils (Rathje et al. 2006). It consists of a flat-end hammer operating within a vertical guide tube. The hammer has a precision accelerometer attached to its end that sends signals to a digital readout unit upon contact with the soil surface. A schematic representation of the CH is pre- sented in Figure 45. The standard hammer has a diameter of 51 mm (2 in.) and weighs 4.5 kg (10 lb). However, CH models with different hammer masses are available. Figure 46 pre- sents photographs of some of these models. The hammer mass used depends on the application. Table 8 presents the various available hammer masses and their applications. Farrag et al. (2005) found that the performance of the 20-kg CH was simi- lar to that of the 10-kg hammer, but larger hammer was less sensitive to small changes in relative compaction. The basic CH system costs approximately $3,000, but the complete sys- tem can cost as much as $20,000. This price is the same for all available hammer masses (Rathje et al. 2006). Principle of Operation The basic principle behind the CH is to obtain a measure of the deceleration of a free-falling mass from a set height onto a soil surface. The standard method for testing using this device is ASTM D5874. According to this method, the ham- mer is to be raised 457 mm (18 in.) after placing the device on a compacted lift. The hammer is then released so that it freely

48 falls within the guide tube. During impact, the accelerometer mounted on the hammer produces an electric pulse, which is converted and displayed on the control unit. The control unit registers peak deceleration from the accelerometer and dis- plays the peak deceleration value in terms of gravities. Four consecutive drops should be performed in the same place, according to the ASTM D5874 standard. The Clegg impact value (CIV) is the largest deceleration measured during the four drops. The ASTM D5874 standard states that the first two blows act as a seating mechanism, with CIVs increasing during the first three drops and remaining generally constant after the fourth. Clegg (1994) proposed equations shown in Table 9 to con- vert CIVs to the Clegg hammer modulus (CHM) for com- monly used Clegg hammers. The equations are derived using double integration of time versus deceleration to determine the deflection, which is then used to compute the elastic modulus based on elastic plate bearing theory. Use of Clegg Hammer in Compaction Control In using the CH for compaction control, it is typically required to specify the target CIV for the soil to be compacted. ASTM D5874 describes three laboratory methods for determining the target CIV. The methods involve measuring the CIV at the optimum moisture content, measuring the CIV at a range of moisture contents, or measuring the CIV at a range of dry den- sities at the optimum moisture content. Each of these methods can use the data obtained from either the standard Proctor or the modified Proctor compaction test. To determine a target CIV for the optimum moisture con- tent, a soil sample is compacted in a Proctor mold at the opti- mum moisture content. The CH is then used to measure the CIV, which represents the minimum required value for field compaction. To determine a target CIV from a range of mois- ture contents, four samples with different moisture contents bracketing the optimum moisture content are compacted in molds at the maximum dry density obtained in the standard or modified Proctor tests. The CIVs are measured for each sample to develop a curve of CIV versus moisture content, and the maximum value is selected as the target CIV in the field. Finally, to set a target CIV from a range of dry densities, four samples are compacted in Proctor molds at the optimum moisture content. Each sample is compacted with a different a. b. FIGURE 46 Clegg hammers: (a) 4.5-kg Clegg hammer, (b) 10/20-kg Clegg hammer (Farrag et al. 2005). FIGURE 45 Clegg impact hammer (modified after ASTM D5874). Hammer Mass (kg) Hammer Diameter (mm) Recommended Applications 0.5 50 Soft turf, sand, golf greens 2.25 50 Natural or synthetic turf (athletic fields) 4.5 50 Preconstructed soils, trench reinstatement, bell holes, foundations 10 130 Flexible pavement, aggregate road beds, trenches 20 130 Reinstatement, bell holes, foundations Source: Mooney et al. (2008). TABLE 8 RECOMMENDED APPLICATION FOR VARIOUS AVAILABLE CLEGG HAMMER MASSES

49 number of blows to produce dry density values ranging from 90% to 100% of the maximum dry density value obtained in standard or modified Proctor tests. The measured CIVs are used to develop a curve of CIV versus relative dry density at the optimum moisture content. The target CIV is selected as the CIV on the developed curve that corresponds to the required percent relative compaction for the site. Repeatability In general, for field use, the coefficient of variation (COV) of CIV is 4% for highly uniform working conditions and 20% for highly variable conditions (ASTM D5874). Mooney et al. (2008) evaluated the repeatability of the CH measurement test and found that the hammer’s precision had an average preci- sion uncertainty of ±4.8%. Rathje et al. (2006) found that the repeatability was medium for both the 10- and 20-kg hammers. Influence Depth Influence depth is the depth in the unbound materials at which the imparted stress by a device becomes negligible. If two or more layers of unbound materials exist within the influence depth of a device, the device measurement will provide a composite value of the two layers rather than the value for the tested layer. Therefore, the determination of the influ- ence depth for test devices is important so that the stiffness/ strength value can be associated with the appropriate lift thickness. Few studies have evaluated the influence depth of the CH. Mooney and Miller (2008) reported that the influence depth ranged between one and one-and-a-half times the hammer diameter and to a maximum of 250 mm (10 in.) for the 10- and 20-kg hammers. White et al. (2007a) found that the influ- ence depth was less than 300 mm (12 in.) for the same size hammers. However, Farrag et al. (2005) reported a lower influence depth of 203 mm (8 in.). Advantages and Limitations The CH is simple to use, requires minimum training, has a standard test procedure (ASTM D5874), and can be outfitted with an integrated GPS system (Farrag et al. 2005; Mooney et al. 2008). In addition, its results can be obtained in a short period of time (less than 60 s) and are not operator depen- dent. Good correlation exists between the CIV and Califor- nia bearing ratio (CBR) values for different types of soils (Aiban and Aurifullah 2007; Fairbrother et al. 2010). Despite these advantages, several limitations of earlier models of the CH were reported in previous studies (Rathje et al. 2006; Mooney et al. 2008). These included poor portability and mobility, particularly with the heavy, 20-kg hammer. In addi- tion, all CH models were found to have weak connections for field use, which significantly affected their durability. The CH was also found to have limited data storage and down- loading capability, which can be an issue when used in large construction projects (Rathje et al. 2006). Farrag et al. (2007) conducted a study to modify and optimize the CH device for soil compaction measurements in the field. The project included physical modifications of the device to reduce its weight and improve its mobility. It also proposed electronic modifications to provide moisture measurement by means of a moisture probe and develop data storage and download- ing capabilities. Although the modified CH model appears to address most of these aforementioned limitations, no studies have been done to report on its efficacy in the field. Mooney et al. (2008) indicated two additional problems with the CH. The first was the inaccuracy of the target CIV obtained by measuring the CIV of a sample compacted in a Proctor mold as a result of boundary effects. The second arose in the testing of soft soils, where the hammer pene- trated the soil so quickly that its handle struck the guide tube. Synthesis of Past Research Studies Indiana Study Kim et al. (2010) reported the results of a study that included performing DCP and CH tests on several road sites within Indi- ana, as well as on clayey soil samples prepared in a test pit and sand samples prepared in a test chamber. The results of this study indicated that the relationship of the CH’s CIV with rela- tive compaction exhibited considerable variability. Based on these findings, the authors suggested that the CH could not be used for compaction control of unbound materials. CHM/H (MPa) = 0.23 (CIV/H)2 CHM/S (MPa) = 0.088 (CIV)2 CHM/M (MPa) = 0.044 (CIV/M)2 CHM/L (MPa) = 0.015 (CIV/L)2 *Note: The 20-kg CH is called the “heavy” CH and thus CIV/H and CHM/H to denote the CIV and the CHM of this mass with its diameter and drop height. Likewise, the 2.25-kg CH is known as the “medium” hammer and thus CIV/M and CHM/M. The 0.5-kg CH is called the “light” CH, so the notation is seen as CIV/L and CHM/L. The 4.5-kg CH is considered the “standard” CH. CIV using the standard CH usually is notated simply as CIV, but CHM/S is useful for distinguishing a Clegg hammer modulus derived using the standard CH. Source: Clegg (1994). TABLE 9 CIV TO MODULUS CONVERSION EQUATIONS

50 Maine Study In a laboratory study in which the CH and LWD were used to test five types of base and subbase aggregates compacted in a test container, Steinart et al. (2005) found that the modu- lus values obtained using the CH measurements were much lower than the LWD moduli. In addition, as shown in Fig- ure 47, a weak correlation was found between the modulus values of the LWD and the CH. The authors attributed the lower CH moduli to the occurrence of a shallow bearing capacity failure caused by the impact of the Clegg Impact Hammer (CIH). They also found that the modulus values determined from the CH’s first drop were less than those obtained from subsequent drops. In addition, the CH moduli tended to increase with each subsequent drop. Texas Study Rathje et al. (2006) evaluated the relationships between CIV and the moisture content and dry density for three types of soils: high-plasticity clay, low-plasticity clay, and well- graded sand. This was done by measuring the CIV for soil samples with a range of moisture contents compacted in Proc- tor molds using standard and modified Proctor compaction efforts, as well as testing soil samples at a constant moisture content and variable dry density. Rathje et al. (2006) indi- cated that for clayey soils, the CIVs were more affected by the moisture content than the dry density. However, for sandy soils the CIVs generally were affected by mois- ture content and dry density such that they increased with increasing dry density. The authors also determined the tar- get CIV, which was chosen as the maximum value obtained over the range of moisture contents tested. For samples compacted using the standard Proctor compaction effort, the target CIVs for high-plasticity clay, low-plasticity clay, and well-graded sand were found to be 7.1, 7.9, and 21, respectively. Virginia Studies Erchul and Meade (1990) studied the correlation between the dry density and CIV obtained from the CH. They concluded that the CIV was a good indicator of the degree of compac- tion for granular materials. The authors found that the use of the CH to estimate dry density values required careful calibration for each material under consideration. In another study, Erchul and Meade (1994) added a penetration scale to the handle of the CH to record the depth to which its ham- mer penetrated the soil. By comparing the CIV and penetra- tion data with density and moisture content measurements obtained using the NDG, the authors developed a graphical acceptance criterion, shown in Figure 48, for utility trench backfill compaction in Chesterfield County, Virginia. FIGURE 47 Comparison between Clegg hammer and LWD modulus (Steinart et al. 2005). FIGURE 48 Acceptance criterion for Clegg hammer (Erchul and Meade 1994).

51 New York State Electric & Gas Corporation Study The New York State Electric & Gas Corporation conducted a field study to compare the CH with the dry density measure- ment obtained using the NDG (Peterson and Wiser 2003). The study involved obtaining 15 measurements at 12 trench backfill sites in Broome County, New York, consisting of crushed rock and gravel. Readings were taken after each lift was compacted to 90% standard Proctor with a tamper. Target CIVs were based on 90% standard Proctor dry unit weight. The study determined that the CH accurately identi- fied the 90% relative compaction for 84% of the measure- ments obtained (Peterson and Wiser 2003). Gas Technology Institute Studies Farrag et al. (2005) reported the results of a study conducted by the Gas Technology Institute, which included testing trenches constructed with various types of unbound materials using 10-kg and 20-kg CHs. The relationships between the 10-kg hammer CIV, 20-kg hammer CIV, and the relative compactions of sand, silty clay, and stone were investigated in that study. Figure 49 shows the relationship between the 10-kg hammer CIVs and relative compaction of sand. Based on the field test results, the authors concluded that the values from both CH models had weak correlations with the relative density for sand and stone-base materials and better correlations in silty clay soil. The CIVs corresponding to 90% relative compaction found in that study are summarized in Table 10. The effect of moisture content on 10-kg hammer CIV was also investigated in the same study for all three types of unbound materials. Figure 50 presents the variation of CIVs with the moisture content of tested sand material. The CIVs increased with the increase in the moisture content to a certain point and then decreased at higher moisture contents. Similar results were obtained for the other tested materials. Farrag et al. (2005) also reported that the maximum moisture content val- ues were not necessarily equal to the optimum moisture values obtained from modified Proctor tests. Finally, the study con- cluded that the performances of the 20-kg and 10-kg hammers were similar. International Studies The CH has been evaluated by numerous international stud- ies, the majority of which looked at the correlation between the CIV and the CBR. In general, good correlations were found between CIV and CBR for different types of unbound materials (Clegg 1980; Mathur and Coghlan 1987; Gulen and McDaniel 1990; Pidwerbesky 1997; Al-Amoudi et al. 2002; Aiban and Aurifullah 2007; Fairbrother et al. 2010). Most researchers found the relationship to be exponential. The first correlation, shown in Eq. 10, was presented by Clegg (1980), which was based on laboratory tests done in soils in Australia. Clegg (1987) used the data collected from laboratory and in situ tests conducted on a wide range of soils in Australia, New Zealand, and the United King- dom to propose a slightly modified correlation, shown in Eq. 11. Al-Amoudi et al. (2002) conducted comprehensive FIGURE 49 The 10-kg hammer CIV compared with relative compaction in sand (Farrag et al. 2005). FIGURE 50 Effect of moisture contents on CIV results in sand (Farrag et al. 2005). Hammer Type Sand Silty Clay Stone-base 10-kg Hammer (CIV) 6 8 14 20-kg Hammer (CIV) 5 6 9 Source: Farrag et al. (2005). TABLE 10 CLEGG HAMMER RESULTS CORRESPONDING TO 90% RELATIVE COMPACTION AT OPTIMUM MOISTURE CONTENT

52 laboratory and field testing programs to evaluate the CIV– CBR correlation and proposed the correlations shown in Table 11 for different soil types. CBR 0.07 CIV (10)2( )= R[ ] )()(= + =CBR 0.24 CIV 1 0.916 (11)2 2 Based on laboratory testing of steel slag and limestone aggregate base materials, Aiban and Aurifullah (2007) pro- posed a slightly different model, shown in Eq. 12, than that proposed by Clegg (1980). More recently, Fairbrother et al. (2010) tested 17 subgrade soil samples that were collected from six locations in the East Cape region of New Zealand. Based on those tests, they proposed Eq. 13 to correlate the CIV to the CBR. It can be noted that Fairbrother et al. (2010) recommended that their equation not be used to estimate the Type of Data Correlations R2 Laboratory CBR = 0.1977 (CIV)1.535 0.810 In situ GM soil CBR = 0.8610 (CIV)1.1360 0.757 SM soil CBR = 1.3577 (CIV)1.0105 0.845 GM and SM soils (combined) CBR = 0.1977 (CIV)1.0115 0.846 Laboratory, in situ and literature data CBR = 0.1691 (CIV) 1.695 0.850 TABLE 11 SUMMARY OF CBR-CIV RELATIONSHIP PROPOSED BY AL-AMOUDI ET AL. (2002) CBR of soft subgrade soils because it will overestimate the CBR strength of soils in that condition. CBR 0.513 CIV 0.94 (12)1.417 2R( )( )= = CBR 0.564 CIV (13)1.144( )= Few studies have compared the CH’s CIV with moduli obtained using other in situ test devices. Whaley (1994) con- ducted a study that included testing base course materials using the Loadman LWD, standard falling weight deflectom- eter (FWD), CH, and Benkelman beam. Figure 51 presents a comparison of measurements obtained from the different devices. Whaley (1994) concluded that poor correlation exists between the CH and the other considered in situ test devices. Pidwerbesky (1997) reached a similar conclusion. FIGURE 51 Comparison between Clegg hammer and other in situ tests (Whaley 1994).

53 SOIL STIFFNESS GAUGE (GEOGAUGE) The soil stiffness gauge, or GeoGauge (Figure 52), measures the in-place stiffness of compacted soil at the rate of about one test per 1.25 min. It weighs about 10 kg (22 lb) and measures 280 mm (11 in.) in diameter and 254 mm (10 in.) in height. The GeoGauge rests on the soil surface by means of a ring–shaped foot. Its annular ring contacts the soil with an outside diameter of 114 mm (4.5 in.), an inside diameter of 89 mm (3.5 in.), and a thickness of 13 mm (0.5 in.) (Lenke et al. 2003). The price reported in previous studies for the GeoGauge ranged between $5,000 and $5,500 (Mooney et al. 2008). The testing procedure of the GeoGauge involves setting it on the test location and giving the device a slight twist to ensure a minimum of 80% contact between the foot and the soil. The manufacturer recommends using a thin layer of sand when 80% contact cannot be achieved. Principle of Operation The principle of operation of the GeoGauge is to generate a very small dynamic force at frequencies of 100 to 196 Hz. In a laboratory study, Sawangsuriya et al. (2002) estimated the force generated by the GeoGauge to be 9 N. The GeoGauge operation includes generating very small displacements to the soil, which is less than 1.27 × 10-6 m (0.0005 in.), at 25 steady state frequencies between 100 and 196 Hz. The stiffness is determined at each frequency, and the average is displayed. The entire process takes about 1.5 min. The GeoGauge is pow- ered by a set of six D-cell batteries. It is designed such that the deflection produced from equipment operating nearby will not affect its measurements because the frequency generated by traffic (at highway speed) is approximately 30 Hz, below the GeoGauge operating frequency (Sawangsuriya et al. 2001). The force applied by the shaker and transferred to the ground is measured by differential displacement across the flexible plate by two velocity sensors (Figure 53). This can be expressed using Eq. 14. At frequencies of operation, the ground-input impedance will be dominantly stiffness con- trolled, such that soil stiffness can be obtained using Eq. 15. F K X X K V Vdr ) )( (= − = − (14)flex 2 1 flex 2 1 where Fdr = force applied by shaker, Kflex = stiffness of the flexible plane, X1 = displacement at rigid plate, FIGURE 52 Soil stiffness gauge or GeoGauge. FIGURE 53 Schematic of the GeoGauge (Humboldt 1998).

54 X2 = displacement at flexible plate, V1 = velocity at rigid plate, and V2 = velocity at flexible plate. K X X X n V V K Knsoil flex flex= = −( )      − ∑ 2 1 1 1 2 1( )     ∑ V n n 1 1 15( ) where n = number of test frequencies, and Ksoil = stiffness of soil. Using velocity measurements eliminates the need for a nonmoving reference for soil displacement and permits accurate measurement of small displacements. It is assumed that GeoGauge response is dominated by the stiffness of the underlying soil. The measured soil stiffness from the GeoGauge can be used to calculate the soil elastic modulus. The static stiff- ness, K, of a rigid annular ring on a linear elastic, homoge- neous, and isotropic half space has the following functional form (Egorov 1965): 1 (16)2K ER v n( ) ( )= − ω where E = modulus of elasticity, V = Poisson’s ratio of the elastic medium, R = the outside radius of the annular ring, and w (n) = a function of the ratio of the inside diameter and the outside diameter of the annular ring. For the ring geometry of the GeoGauge, the parameter w (n) is equal to 0.565, thus 1.77 1 (17)2K ER v( )= − Based on Eq. 17, the GeoGauge stiffness could be converted to an elastic stiffness modulus using the equation proposed by CA Consulting Engineers, as follows (Eq. 18): 1 1.77 (18) 2 E H vRG SG ( ) = − where EG = the elastic stiffness modulus (MPa), HSG = the GeoGauge stiffness reading (MN/m), and R = the radius of the GeoGauge foot [57.15 mm (2.25 in.)]. For a Poisson’s ratio of 0.35, a factor of approximately 8.67 can be used to convert the GeoGauge stiffness (in MN/m) to a stiffness modulus (in MPa). It is recommended that the GeoGauge be used only for materials with stiffness to 23 MN/m because it may lose accuracy when measuring stiffness greater than that value (Chen et al. 2000). The GeoGauge manufacturer also has suggested that the dry density of compacted soils can be determined from GeoGauge stiffness using Eq. 19, which was developed based on the work of Hryciw and Thomann (1993). In this equa- tion, the calibration factor “C” should first be determined for tested geomaterial. This is done by measuring GeoGauge stiffness (K) along with the dry density of the geomaterial to be tested and solving Eq. 20. Several studies indicated that the GeoGauge performed poorly when used to determine dry density. 1 1.2 0.3 (19)0.5Cm K D oρ = ρ + −  1 1.2 0.3 (20) 2 C K m o D =   ρ ρ −        +       where rD = the dry density of the soil; ro = the ideal, void free density; and m = moisture content. Repeatability Several previous studies have indicated that the GeoGauge had similar or better repeatability than other in situ test devices. Maher et al. (2002) reported that the GeoGauge had excellent repeatability when conducting consecutive measurements on different soil types. Abu-Farsakh et al. (2004) found that the GeoGauge’s COV was between 0.2% and 11.38% for field test sections and ranged from 2.3% to 38.8% for tests con- ducted in laboratory test sections. Von Quintus et al. (2008) indicated that when testing seven types of soils at seven test sites, the COVs of GeoGauge measurements ranged from 7.1% to 20.1%. In another field study, Hossain and Apeagyei (2010) found that the GeoGauge had lower spatial variabil- ity than did the LWD and DCP, but the reported COV for GeoGauge moduli was 8% to 42%. As part of tests conducted in Phase I of the NCHRP 10-48 project, Nazarian (2012) reported that the COV of GeoGauge measurement was less than 10%. The precision of GeoGauge measurement on fine-grained soils was reported to be less than 2% and on coarse-grained soils and crushed aggregate less than 5%. The repeatability of the GeoGauge was also evaluated in a field study con- ducted at the Louisiana Transportation Research Center as part of a FHWA study [SPR-2(212)] for the validation of the seating procedure for the GeoGauge. Fifty-four GeoGauge measurements were taken at each test location, and the COV calculated for all measurements made. The COV for mea- surements made by all GeoGauges ranged from 6.1% to

55 9.5%. Finally, some studies have reported that the GeoGauge results are extremely inconsistent and highly dependent on the seating procedures and the operator (Bloomquist et al. 2003; Mooney et al. 2008). Influence Depth Several studies were performed to determine the zone of influ- ence of the GeoGauge. Nazzal (2003) used two test boxes, one box containing compacted clay and another compacted Florolite (plaster of Paris), to determine zone of influence. The average zone of influence of the GeoGauge was found to be 190 to 203 mm (7.5 to 8 in.). Sawangsuriay et al. (2002) found a zone of influence of 127 to 254 mm (5 to 10 in.) for the GeoGauge using cubic boxes filled with medium sand, crushed lime rock, and a mixture of plastic beads with sand. Maher et al. (2002) reported a similar influence zone for the GeoGauge. Advantages and Limitations The GeoGauge test is simple with minimal training required to perform it. In addition, it is fast (75 s per test) and has a well-defined test specification (ASTM D6758). The GeoGauge device also has good portability, durability, data storage, and download capabilities. Previous studies have reported some of the GeoGauge’s limitations. First, its reading is sensi- tive to the stiffness of the top 2 in. of the tested soil layer, as well as to the seating procedure (Bloomquist et al. 2003; Farrag et al. 2005). Furthermore, previous studies reported that often there was difficulty in achieving good contact between the GeoGauge ring and the tested soil (Simmons 2000; Ellis and Bloomquist 2003; Miller and Mallick 2003). Simmons (2000) also found that the use of leveling sand for surface preparation that is recommended by the manufacturer can significantly affect GeoGauge measurements. Another limitation of the GeoGauge is the very small load that it applies, which does not represent the stress levels actu- ally encountered in the field as a result of traffic. Therefore, the GeoGauge modulus must be corrected to account for design loads. In addition, GeoGauge measurement has been found to be very sensitive to changes in moisture content (Nazzal 2003). Finally, Miller and Mallick (2003) raised con- cerns about the GeoGauge malfunctioning owing to vibra- tions from passing vehicles, such as compaction equipment or trains. Synthesis of Past Research Studies Florida Study Bloomquist et al. (2003) reported the results of a study that evaluated the effectiveness of GeoGauge as a tool for com- paction control of pavement base and subgrade materials. The GeoGauge did not have definitive correlations with dry density, moisture content, or resilient modulus. Furthermore, the repeatability and precision of the GeoGauge was found to be largely dependent on the condition of the soil surface as well as the placement and operation procedure. The research- ers attempted to enhance the design of the GeoGauge by developing a new handle to provide a uniform seating of GeoGauge on the soil. The results indicated that the repeat- ability of the GeoGauge was significantly improved when it was seated on the soil by twisting it with the newly devel- oped handle. Finally, Bloomquist et al. (2003) found that the GeoGauge stiffness value tends to increase as the frequency increases. Therefore, they recommended that certain input frequency ranges be used during testing to reduce the vari- ability in GeoGauge readings. Hawaii Studies Pu (2002) evaluated the relationship between GeoGauge stiff- ness with moisture content, dry unit weight, and CBR. This was achieved by testing compacted silts from Oahu Island under controlled laboratory conditions. Results showed no direct relationship between GeoGauge stiffness and dry unit weight, because a GeoGauge stiffness value can correspond to dif- ferent values of dry unit weight depending on the moisture content. Pu (2002) derived a relationship between GeoGauge stiffness, dry unit weight, and moisture content. However, this relationship requires detailed information on the soil water characteristic curves. The author concluded that the GeoGauge could provide an alternative method for compaction control that used stiffness instead of dry unit weight. However, he indicated that the soil shrink/swell potential is not optimized if stiffness is used. Therefore this issue needs to be addressed before stiffness-based compaction control specification is implemented. Finally, Pu (2002) indicated that no direct relationship existed between GeoGauge measurements and the soaked CBR because the GeoGauge provided a measure of stiffness at a much smaller displacement than that encountered during CBR testing (2.5 to 5 mm). Ooi et al. (2010) conducted GeoGauge and LWD tests on recycled concrete aggregate (RCA) and reclaimed asphalt pavement (RAP) lifts 6-in. thick compacted in bins 3 ft in diameter. The results showed that the LWD consistently pro- vided higher moduli than did the GeoGauge. In addition, the LWD had better repeatability. Finally, the authors suggested that it is important to consider the zone of influence when interpreting LWD and GeoGauge moduli. Louisiana Studies Abu-Farsakh et al. (2004) conducted a comprehensive study to evaluate the use of the GeoGauge, DCP, and LWD to reliably measure the stiffness/strength characteristics of unbound materials for application in the quality control/

56 quality assurance (QC/QA) procedures during and after construction of pavement layers and embankments. The study included conducting GeoGauge, DCP, LWD, standard FWD, and static plate loading (PLT) tests on different base course materials and subgrade soils in several pavement sections at three project sites in Louisiana as well as at the Louisiana Department of Transportation and Development Accelerated Load Facility (ALF). In addition, tests were performed on sections constructed in two laboratory boxes measuring 1.5 × 0.91 × 0.76 m (5 × 3 × 2.5 ft). The CBR laboratory tests were also conducted on samples collected during the testing of different sections. Abu-Farsakh et al. (2004) found that the GeoGauge was the most user-friendly tool among the three devices evaluated in this study because it was durable, easy to operate, and provided rapid results. Based on the results of this study, the following correlations were found between GeoGauge and the two standard in situ tests (FWD and PLT): M E E R FWD G G )( )(= − + < < = 20.07 1.17 for 40.8 MPa 194.4 MPa 0.81 (21)2 E E E R PLT i G G )( )(= − + < < = )( 75.58 1.52 for 40.8 MPa 194.4 MPa 0.87 (22)2 E E E R PLT R G G )( )(= − + < < = )( 65.37 1.50 for 40.8 MPa 194.4 MPa 0.90 (23) 2 2 where MFWD = FWD back-calculated modulus (MPa), EPLT (i) = initial moduli from the PLT (MPa), EPLT(R2) = reloading moduli from the PLT (MPa), and EG = GeoGauge modulus (MPa). Abu-Farsakh et al. (2004) also proposed the regression model shown in Eq. 24 to correlate the GeoGauge moduli with the CBR values obtained by testing soil samples col- lected from field test sections. All samples were prepared in accordance with ASTM D1883-99 without soaking them to mimic the field conditions. E E R G G )( )(= − < < = CBR 0.00392 5.75 for 40.8 MPa 184.11 MPa 0.84 (24) 2 2 Zhang et al. (2004) evaluated the use of the GeoGauge for controlling trench backfill construction. Three trenches were excavated with dimensions of 1.3 × 5 × 1 m (4 × 15 × 3 ft). Each trench consisted of three layers, each with a thickness 300 mm (12 in.). Each trench was then divided into three equal sections compacted at different compaction efforts: light, mod- erate, and heavy. As expected, results indicated that both the dry density and GeoGauge modulus increased with increas- ing compactive effort. However, this depended on soil type as well as moisture content. In another study, Mohammad et al. (2009) conducted resil- ient modulus (Mr) laboratory experiments on soil samples collected from sections tested in the study by Abu-Farsakh et al. (2004). Based on the results of the conducted tests, Mohammad et al. (2009) proposed two models (shown in Eqs. 25 and 26) to predict the laboratory-measured Mr from the GeoGauge modulus. Although the first model directly related the GeoGauge modulus to the Mr value, the second model pre- dicted the Mr value based on the GeoGauge modulus as well as the moisture content of the tested soils. M E Rr G46.48 0.01 0.59 (25)1.54 2 )(= + = M E w R r G )( ) ) ( (= − + + = 13.94 0.0397 601.08 1 0.72 (26) 0.8 0.78 2 where Mr = resilient modulus (MPa), EG = modulus from GeoGauge test (MPa), and w = moisture content (%). Minnesota Studies Siekmeier et al. (2000) compared the GeoGauge to other in situ test devices, such as the Loadman LWD and the standard FWD. The results of this study showed that the GeoGauge modulus was less than those measured by other in situ test devices. This was attributed to the lower stress imposed on the soil by the GeoGauge (0.02 to 0.03 MPa) compared with that imposed by the FWD and LWD (0.7 to 0.9 MPa). In addition, the resilient modulus values of the tested soils measured in the laboratory were found to be approximately twice those obtained by the GeoGauge. This suggested that the stress levels used during the labo- ratory testing may be much higher than those imposed by GeoGauge. Petersen and Peterson (2006) reported the results of an intelligent compaction demonstration project in which the GeoGauge and the LWD were used to test the final lift of a 914-mm (3-ft) subcut consisting of a select granular bor- row material. The material was compacted using a vibratory compaction roller outfitted with intelligent compaction (IC) technologies. The GeoGauge was conducted at 42 points along the project. Results showed a poor correlation between GeoGauge modulus and the IC roller measurements when the comparison was done on a point-by-point basis. This was attributed to the relatively shallow depth of influence of the GeoGauge and the soil’s heterogeneity. However, a relatively good correlation was obtained between the GeoGauge and the LWD. In addition, the authors found that the GeoGauge was easy to use by a single individual, and it provided repeat- able measurements when properly seated.

57 New Jersey Study Maher et al. (2002) conducted field and laboratory inves- tigations to evaluate the suitability of the GeoGauge for soil compaction control and dry density measurement. The field component included performing GeoGauge and nuclear density gauge tests at 400 points during placement and com- paction of two embankments composed of Portland cement- stabilized dredge sediments. Approximately 50 points in the first embankment were used for the calibration. The labora- tory investigation involved testing three types of subgrade soils and one subbase aggregate material compacted in a 55-gallon steel drum cut to a maximum height of 610 mm (24 in.). Field work results indicated that the GeoGauge could indeed be used to estimate the dry density of soils if proper cal- ibration factors were determined for the tested soil. Laboratory results indicated that the GeoGauge had the potential to deter- mine the resilient modulus of soils; however, calibration to the different applied stress conditions was needed for validation. New Mexico Study Lenke et al. (2003) evaluated the use of the GeoGauge for com- paction control of pavement materials. Their results indicated that the GeoGauge was able to detect the increase in soil stiff- ness with compaction by roller passes (Figure 54). However, the authors found that their attempts to determine a field target value for the GeoGauge in the laboratory using modified Proc- tor molds were not successful because the GeoGauge annular foot size is comparable with that of the mold. Lenke et al. (2003) indicated that without being able to develop a laboratory- determined target value for stiffness in the field, compac- tion control specifications that use GeoGauge should include careful control of the moisture content of compacted soils. In addition, they suggested the use of test strips to determine the GeoGauge compaction control parameters for a given soil. Texas Study Chen et al. (1999) used the GeoGauge, traditional FWD, dirt seismic pavement analyzer (D-SPA), and Olson spectral analysis of surface waves (SASW) to measure the stiffness of base course materials at eight different locations. Results indicated that the modulus measured with the FWD was higher than that measured with the GeoGauge. The authors sug- gested a general relationship between GeoGauge stiffness and the FWD back-calculated modulus as follows: 37.65 261.96 (27)M HFWD SG= − where MFWD = back-calculated FWD modulus (MPa), and HSG = GeoGauge stiffness reading (MN/m). LIGHT WEIGHT DEFLECTOMETER The light weight deflectometer (LWD) is a portable fall- ing weight deflectometer that consists of a falling mass and a displacement-measuring sensor attached at the center FIGURE 54 Variation of the GeoGauge with roller passes (Lenke et al. 2003).

58 FIGURE 55 Schematic drawing of LWD showing various components of the equipment (Kim et al. 2010). Device Plate Diameter (mm) Plate Thickness (mm) Falling Weight (kg) Maximum Applied Force (kN) Load Cell Total Load Pulse (ms) Type of Buffers Deflection Transducer Type Location Measuring Range (mm) Zorn 100, 150, 200, 300 124, 45, 28, 20 10, 15 7.07 No 18 ± 2 Steel spring Accelerometer Plate 0.2–30 (±0.02) Keros 150, 200, 300 20 10, 15, 20 15 Yes 15–30 Rubber (conical shape) Velocity Ground 0–2.2 (±0.002) Dynatest 3031 100, 150, 200, 300 20 10, 15, 20 15 Yes 15–30 Rubber (flat) Velocity Ground 0–2.2 (±0.002) Prima 100, 200, 300 20 10, 20 15 Yes 15–20 Rubber (conical shape) Velocity Ground 0–2.2 (±0.002) Loadman 110, 132, 200, 300 Unknown 10 17.6 No 25–30 Rubber Accelerometer Plate Unknown ELE 300 Unknown 10 Unknown No Unknown Unknown Velocity Plate Unknown TFT 200, 300 Unknown 10 8.5 Yes 15–25 Rubber Velocity Ground Unknown CSM 200, 300 Unknown 10 8.8 Yes 15–20 Urethane Velocity Plate Unknown Notes: Zorn Light Drop Weight Tester ZFG2000 by Gerhard Zorn, Germany; Keros Portable FWD and Dynatest 3031 by Dynatest, Denmark; Prima 100 Light Weight Deflectometer by Carl Bro Pavement Consultants, Denmark; Loadman by AL-Engineering, Oy, Finland; Light Drop Weight Tester by ELE; TRL Foundation Tester (TFT) is a working prototype at the Transport Research Laboratory, United Kingdom; Colorado School of Mines (CSM) LWD device. After White et al. (2009b). TABLE 12 COMPARISON BETWEEN DIFFERENT LWD DEVICES of a loading plate. Figure 55 provides a schematic repre- sentation of the LWD, including its various components. There are several types of LWD on the market that have been evaluated in previous studies, including the Dynatest, Keros, German dynamic plate (GDP), Prima 100, Trans- port Research Laboratory (prototype) Foundation Tester (TFT), and Zorn. These devices exhibit many similarities in their mechanics of operation, although differences in design and mode of operation lead to variations in the measured results. Table 12 provides a comparison of the different LWD devices. Figures 56 and 57 present photographs of two types of LWD. Principles of Operation The LWD test is performed by releasing the falling weight from a standard height onto the loading plate using the top fix-and-release mechanism. An impulse load is imposed on the compacted material through the plate. The resulting central deflection of the loading plate is obtained either by integrating the velocity measurements taken from a veloc- ity transducer or by double-integrating the acceleration data taken from an accelerometer. The expression used to calcu- late LWD modulus is similar to the one used to calculate the surface modulus of a layered system having a homoge- neous properties, assuming constant loading on an elastic half space (Boussinesq elastic half space). This expression is shown in Eq. 28. Currently, ASTM E2583-07 is the stan- dard method for conducting LWD tests. Some state DOTs (Indiana and Minnesota) have developed standard test pro- tocols for the LWD. E v R ALWD c )( = − σ × × δ 1 (28) 2

59 where s = the applied stress, R = the loading plate radius, v = Poisson’s ratio (usually set in the range of 0.3 to 0.45 depending on test material type), dc = central peak deflection, and A = plate rigidity factor: default is 2 for a flexible plate, p/2 for a rigid plate. Factors Influencing the LWD Modulus A number of factors may influence the measured LWD mod- ulus, including falling mass, drop height, plate size, plate contact stress, type and location of deflection transducer, usage of load transducer, loading rate, and buffer stiffness (Fleming 2001; White et al. 2007a). Contrasting information is available in the literature on the effect of plate size on mea- sured LWD modulus. Fleming et al. (2007) studied the effect of the plate size and drop weight on stiffness. Their results indicated that for the 15- or 20-kg drop mass, the modulus did not change significantly with different plate diameters. However, Chaddock and Brown (1995) demonstrated that using the TFT LWD with a 200-mm plate resulted in a mod- ulus that was approximately 1.3 to 1.5 times greater than that with a 300-mm plate. Furthermore, based on field stud- ies conducted using LWD, Deng-Fong et al. (2006) found that the LWD modulus measured using a 100-mm plate was about 1.5 times greater than that from a 300-mm plate. Lin et al. (2006) also concluded that the size of the loading plate was a significant factor affecting LWD modulus. They indi- cated that the use of an inappropriate loading plate could affect the measurements and the modulus calculation. Lin et al. (2006) also evaluated the effect of drop heights, con- cluding that different drop heights had very little effect on stiffness. To get consistent and comparable results, research- ers have suggested using a LWD with the same mass, drop height, and plate size. Davich et al. (2006) recommend using a LWD with a mass of 10 kg, drop height of 500 mm, and plate diameter of 200 mm. They suggested that this combina- tion resulted in the test volume extending to the bottom of a common lift. Differences in type and location of sensors used in LWD devices can also lead to variations in measured LWD mod- ulus. White et al. (2007a) compared the subgrade moduli measured using two LWD devices: the Zorn (Model ZFG 2000) and the Keros. The moduli measured with the Keros were found to be 1.75 to 2.2 times greater. The researchers attributed the differences in the measured modulus values to the different methods used to measure deflections in both devices. Although the Keros measures deflections on the FIGURE 56 Dynatest 3031 LWD (Dynatest 2013). FIGURE 57 Zorn LWD (Zorn Instruments 2013).

60 ground with a geophone, the Zorn uses accelerometers that measure plate deflection, which is expected to measure larger deflections. Some LWD devices (e.g., Zorn) assume a constant applied force based on calibration tests performed on a concrete sur- face, whereas others (e.g., Prima 100 and Keros) use a load cell to measure the actual applied load during the test. Previ- ous studies have concluded that the assumption of constant applied force does not significantly affect the measured mod- ulus when using the LWD to test relatively stiff compacted layers (Brandl et al. 2003; White et al. 2007a). Some studies suggested that the spring stiffness of the buf- fer placed between the drop weight and the contact plate con- trols the loading rate and thus can affect the measured LWD modulus. Adam and Kopf (2002) found that the applied load pulse varied by about 30% with a change in rubber buffer temperature from 0°C to 30°C; it remained more constant, however, using a steel-spring buffer. This might explain why Germany has prohibited the use of rubber buffers (White et al. 2007a). Finally, Fleming (2000) found that a compara- tively lower stiffness buffer provided more efficient results. Repeatability The reliability of LWD measurement is significantly influ- enced by its repeatability. Petersen et al. (2007) and Hossain and Apeagyei (2010) reported a relatively high COV, ranging from 22% to 77% for LWD-measured modulus when test- ing various types of unbound materials. Von Quintus et al. (2008) reported a COV between 13.9% and 77.3% for differ- ent types of LWD devices. Nazzal et al. (2007) evaluated the repeatability of the LWD by using the COV of five measure- ments taken at the same testing point. Figure 58 shows the COV with the corresponding average LWD elastic moduli measured in that study. The COV of the LWD measurements ranged from 2.1% to 28.1%. The general trend for the points in Figure 58 indicates that COV values decreased as LWD elastic moduli increased. Nazzal et al. (2007) indicated that during field testing it was difficult to conduct LWD tests on very weak subgrades because of uneven surfaces that caused tilting of the loading plate. Another reason for high COV values in weak subgrade soils was that those soils exhib- ited significant permanent deformation under the LWD test. Similar findings were reached by other researchers (e.g. Fleming 2000; George 2006; Fleming et al. 2007). Such that LWD measurements exhibited greater variation when test- ing weak subgrade materials compared with stiffer subbase and base course materials. Fleming et al. (2009) reported that the range of COV of LWD measurements varied between 25% and 60% for fine-grained subgrades owing to variation in moisture content. In addition, for granular capping (sub- base) layers, the COV ranged from 10% to 40%, with higher values observed on very wet sites. For highly specified, well- graded, crushed aggregate base materials, the COV of LWD measurements typically was less than 15%. Several factors affecting the variability in the moduli measured with the LWD have been reported in the liter- ature. They include (1) the number of load drops, (2) the quality of the load and deflection curve, and (3) the level of contact between loading plate and tested layer surface (White et al. 2007a; Fleming et al. 2007; Ooi et al. 2010). Steinart et al. (2005) studied the influence of the number of load drops on the measured LWD modulus and found that the measure- ments from the first drop typically were smaller than those derived from subsequent ones, as shown in Figure 59. There- 0 50 100 150 200 250 300 350 400 ELFWD (MPa) 0 4 8 12 16 20 24 28 32 C V % FIGURE 58 Cv variation with LWD modulus (Nazzal et al. 2007). FIGURE 59 Effect of consecutive drops on composite modulus values (Steinart et al. 2005).

61 fore, they recommended that the first value be excluded in calculating the average LWD moduli value. Davich et al. (2006) suggested using three LWD seating drops followed by three drops at each test location to produce consistent LWD data. George (2006) recommended that two seating loads be applied at each station, followed by four or more load drops of 1,730 lb. In addition, George (2006) suggested that the LWD load-deflection history should be checked continuously for inconsistencies. Fleming et al. (2007) rec- ommended that the deflection-time history of each drop be assessed to determine the quality of the LWD results. The authors demonstrated (Figure 60) that an ideal test should have no deflection at time zero, but then should increase to a peak, followed by a decrease. At the end of the pulse, no deflection should occur. The authors also illustrated a pos- sible deflection-time (pulse) history other than the ideal one. For example, Figure 61 shows the deflection at the end of the pulse moving in the opposite direction instead of returning to zero. This typically happens when the instrument bounces off the ground upon impact. This type of deflection-time history is also possible if the tested material contains excess water. The effect of successive drops at the same spot on deformation-time histories is shown in Figure 62. Both the maximum and final deflections decrease as the number of repetitions increase. Figures 60 through 62 provide a guide for determining acceptable readings. Previous studies have recommended that the LWD be conducted on a uniform surface to ensure optimal contact between the LWD loading plate and the tested material. For example, Lin et al. (2006) found that the repeatability of the LWD was very good only if there was an even contact sur- face. Higher variability was observed for uneven surfaces (e.g., coarse gravel). Remedies for uneven surfaces include using moist sand, removing as much as 102 mm (4 in.) of the compacted material before testing, and limiting testing to layers with a gradient of less than about 5%. Influence Depth Nazzal et al. (2007) evaluated the influence depth of the LWD by conducting laboratory tests inside two test boxes (1,824 × 912 × 912 mm). To clearly define the influence zone for the LWD, stiff soil was constructed on top of soft soil and vice versa. The results indicated that the influence depth for the LWD ranges between 270 and 280 mm (10.6 and FIGURE 60 High-quality LWD reading (Fleming et al. 2007). FIGURE 61 Example for high rebound LWD reading (Fleming et al. 2007).

62 11 in.), which was about 1.5 times the diameter of the loading plate. Brandl et al. (2003) reached similar findings. However, Fleming et al. (2007) and Siekmeier et al. (2000) reported a lower influence zone equal to the diameter of the loading plate. Based on analysis of in situ strain data, Mooney et al. (2009) found that the LWD influence depth ranged from 0.9 to 1.1 of the loading plate diameter depending on soil type. Advantages and Limitations Several advantages of the LWD were reported in the literature. The setup and test times for LWD are relatively short (Sebesta et al. 2006; Siekmeier et al. 2009). In addition, the LWD mea- sures the modulus value of tested pavement materials, which can be directly used as input in the pavement design. With additional sensors, the LWD can distinguish stiffness values between pavement layers. Siekmeier et al. (2009) indicated that the LWD could accurately test more material types, such as unbound materials with large aggregates, than could the standard density-based approach. In addition, LWD testing is safer because the field inspector is able to remain standing and visible during most of the testing process (Davich et al. 2006). The LWD’s main limitation is its high variability. Hossain and Apeagyei (2010) reported high variability in measured LWD modulus for the same material tested with different LWD devices. Other studies reported poor repeatability when testing weak cohesive materials or layers with uneven surfaces (Nazzal 2003). Petersen et al. (2007) indicated that the LWD tended to move during testing, which affected the reliability of the test results. They recommended using a two-buffer configuration when testing using the LWD to increase the dampening of the impact load and to limit the movement of the machine during testing. To provide uni- form loading and reduce machine movement, the authors recommended that a smooth and level test area be selected so that a good contact exists between the loading plate and the test surface. Another problem, reported during an interview with the Minnesota DOT, is the difficulty encountered when using the LWD in large projects as a result of its relatively heavy weight. Hossain and Apeagyei (2010) also found that the effect of the moisture content on the LWD-measured modu- lus was much more significant compared with the moduli measured with other in situ test devices. Finally, there are some concerns about the effectiveness of the LWD in testing layered systems. This concern is mainly attributable to the possibility that the LWD’s zone of influ- ence may extend beyond the thickness of the tested layer. Sebesta et al. (2006) recommended using a three-sensor sys- tem that reduces the setup time and measures modulus values of the multiple pavement layers. Lin et al. (2006) indicated that with the three sensors the moduli for three layers can be computed based on the measured deflections and distances to the load. In this case, the modulus computed from deflec- tions further away from the load represents the deeper lay- ers. However, Steinart et al. (2005) suggested that until a program is developed to incorporate the deflections from all three sensors simultaneously into a back-calculation routine, the additional sensors will not be useful. Synthesis of Past Research Studies Kansas Study Petersen et al. (2007) investigated the use of the LWD as a tool for compaction quality control of embankment soil. LWD and FWD tests were conducted on different types of soils in nine Kansas DOT embankment projects. Density and moisture measurements were taken at selected test locations, and resilient modulus tests were conducted on soil samples FIGURE 62 Effect of increasing number of blows on LWD reading (Fleming et al. 2007).

63 obtained during field testing and prepared in the laboratory at varying density and moisture contents. Figure 63 presents the results of the LWD and FWD tests conducted in that study. It is clear that the LWD modulus was very close to that measured by the standard FWD. Petersen et al. (2007) indicated that they failed to develop a model that relates the LWD modulus to the laboratory-determined resilient modulus. They attributed that to the differences in the state of stress as well as the dry density and moisture content conditions of soil tested in the field as compared with that in laboratory. The authors also suggested that the moisture content of the soil in the field may vary within the testing area owing to desiccation of the surface layer that occurs between the end of compaction and the onset of stiffness testing. Finally, they found that the high degree of spatial vari- ability obtained for the LWD moduli prevented the develop- ment of a quality control procedure based on a control test strip. Louisiana Studies Abu-Farsakh et al. (2004) conducted field and laboratory testing programs to evaluate the effectiveness of the LWD in measuring the stiffness properties of different types of geo- material for application in the QC/QA procedures during and after the construction of pavement layers and embankments. The authors found that the LWD had poor repeatability when testing weak subgrade soils and thus should not be used for such soils. In addition, they indicated that for pavement lay- ers less than 305 mm (12 in.) thick, the LWD measurement might not reflect the true modulus value of the tested layer but rather a composite modulus for the multiple layers below. In this case, the authors recommended that the LWD modulus be back-calculated. Based on the results of regression analy- sis conducted on the data obtained in this study, Abu-Farsakh et al. (2004) found the following correlation between LWD and FWD back-calculated resilient moduli, MFWD, PLT initial and reloaded modulus, and CBR: M E E R FWD LFWD LFWD 0.97 for 12.5 MPa 865 MPa 0.94 (29)2 )( )(= < < = E E E R PLT i LFWD LFWD )( )(= + < < = )( 22 0.7 for 12.5 MPa 865 MPa 0.92 (30)2 E E E R PLT R LFWD LFWD )( )(= + < < = )( 20.9 0.69 for 12.5 MPa 865 MPa 0.94 (31) 2 2 E E R LFWD LFWD ( ) ( )= + < < = CBR –14.0 0.66 for 12.5 MPa 174.5 MPa 0.83 (32)2 In another study, Mohammad et al. (2009) reported the model in Eq. 33 directly predicts the laboratory measured Mr from the LWD modulus. To enhance the prediction, the moisture content was included as a variable in the Mr regres- sion model (Eq. 34). It is noted that the moisture content was chosen based on stepwise selection analysis that included various physical properties of the tested unbound materials. M E Rr LFWD )(= × =27.75 0.54 (33)0.18 2 M E w R r LFWD )( ) ) ( (= + + = 11.23 12.64 242.32 1 0.7 (34) 0.2 2 where Mr = resilient modulus (MPa), ELFWD = modulus from LWD test (MPa), and w = moisture content (%). Mississippi Study George (2006) reported the results of a study to investigate the effectiveness of the LWD in testing subgrade soil. LWD, standard FWD, and nuclear density gauge measurements were obtained at 13 as-built subgrade sections reflecting typical subgrade soils in Mississippi. Resilient modulus and other routine laboratory tests were conducted on soil samples collected from the sec- tions. The author concluded that the LWD is a viable device for characterizing subgrade soil provided that the imposed stress level is within the linear elastic range of the tested soil. He pro- posed Eq. 35 to relate the LWD modulus to the in-place density and moisture content of the tested soil. In addition, he devel- oped a model (Eq. 36) to predict the laboratory-determined resilient modulus of soil compacted at 95% relative compac- tion from ELWD and dry density, moisture content, and soil index properties. Because of concerns raised by the Mississippi DOT engineers about the availability of soil index properties, namely PI and P200, George (2006) developed another model without the PI/P200 term (Eq. 37). Finally, based on the field test results, the author found the correlation shown in Eq. 38 between LWD and FWD back-calculated moduli. E w RLWD f o f) )( ( )(= =) )( ( −109,988 D 0.83 (35) 5.544 0.594 2 FIGURE 63 Relationship between LWD and FWD moduli.

64 E M D M PI P R LWD R f o f o ( ) = − + − − = ( ) ( )2.30 3.860 0.316 0.635 0.83 (36) 95 200 2 E M D M R LWD R f f o ( )= − + − =( ) ( )3.907 5.435 0.370 0.70 (37) 95 95 2 M E E R FWD LWD LWD )( = < < = 1.09 , 2240 psi 30740 psi 0.64 (38)2 where ELWD = measured LWD elastic modulus (psi), D(f/o) = ratio of field unit weight to unit weight at optimum moisture, w(f) = field moisture (%), D(f/95) = ratio of field unit weight to unit weight at 95% compaction, and M(f/o) = ratio of field moisture to optimum moisture. Maine Study Steinart et al. (2005) studied the effectiveness of the LWD as a tool for compaction control of subgrade and base materials using field and laboratory tests. The laboratory component of the study included compacting five types of base and subbase aggregates in a 1.8 × 1.8 × 0.9 m (6 × 6 × 3 ft) deep test container using 152 mm (6 in.) lifts. The compacted materials were tested using various in situ test devices: the LWD, CH, DCP, and NDG. The field component included testing two subgrade soils, sand, two base aggregates, and one reclaimed stabilized base product using the LWD and NDG. In general, the laboratory and field test results showed that the LWD modulus generally increased with increases in the percent compaction. However, whereas the laboratory tests showed poor correlation between LWD modu- lus and percent compaction, the field test sites where granular base materials were tested yielded a good correlation. In addi- tion, the results of multivariable regression analyses conducted on the field test data to predict the LWD modulus as a function of percent compaction and moisture content yielded the model shown in Eq. 39. The obtained model had a relatively high R2, which suggested that a strong correlation existed between the LWD modulus and percent compaction and moisture content. However, the authors indicated that the moisture content values for the tested field sites were all on the dry side of optimum, which may limit the significance of this conclusion. 411.26 5.454 – 2.757 0.823 (39)2 E PC RWC R LWD ( ) ( ) ( )= − + = where ELWD = LWD composite modulus, PC = percent compaction, and RWC = relative water content. Minnesota Studies An earlier work done by Minnesota DOT was a study reported by Siekmeier et al. (2000) in which Loadman LWD, DCP, GeoGauge, and traditional FWD were used to test granular base materials at several construction sites in Minnesota. For each test location, five LWD tests were performed and the last three measurements were averaged. Laboratory resilient modulus tests also were conducted on cores obtained during field testing. Figure 64 presents the results by the different in situ devices. The LWD modulus had a trend similar to that of the FWD back-calculated modulus but a different magnitude. The authors attributed the observed differences to the variation in the stress conditions imposed by each of these two devices. As shown in Figure 64, there was little agreement between relative compaction and measured in situ moduli. The authors explained that it was not realistic to know the Proctor maxi- mum density for every soil type found at a construction site. They suggested that compaction tests could be compared with in situ modulus tests only when the material is uniform with respect to a single maximum Proctor density (Siekmeier et al. 2000). Hoffmann et al. (2003) indicated that prediction of the LWD modulus based on load and peak deflections could result in inaccurate modulus values; therefore, to improve prediction, they proposed a spectral-based procedure to ana- lyze LWD data. Davich et al. (2006) presented the results of a study con- ducted by the Minnesota DOT’s Office of Materials to provide data needed for developing LWD compaction control specifi- cations. LWD and moisture meter tests were conducted on FIGURE 64 Moduli compared with location for granular base material (Siekmeier et al. 2000).

65 samples of three types of granular materials, which were compacted inside an open-topped, steel cylinder (half a 55-gallon steel drum) using a procedure similar to that of the standard Proctor test. The results of this study showed that the LWD provided a level of accuracy similar to that of DCP testing. However, the LWD has an advantage over the DCP because it directly measures quantities that character- ize the pavement layers’ mechanical response during traffic loading, such as force and displacement. Furthermore, it is nondestructive and requires less inspector effort than does DCP testing. The authors recommended that LWD plate size and falling mass drop height should be standardized to obtain consistent and reliable data. As part of the Minnesota DOT’s efforts to evaluate and implement intelligent compaction technology and other in situ tests into earthwork construction practice, White et al. (2007a) presented the results of a field study in which two types of LWD devices (Zorn and Keros) with different plate diameters were used to test subgrade soils and base course layers at con- struction sites in Minnesota. In addition, this research included conducting resilient modulus tests on Shelby tube samples obtained from the tested subgrade soils at the locations of LWD tests. The results of this study showed that the LWD modulus measured using the Keros device was on average 1.9 to 2.2 times greater than that measured with the Zorn. The authors attributed the differences in measured modulus val- ues between the two devices to the Zorn measuring approxi- mately 1.5 times greater deflection than did the Keros for the same plate diameter. The authors also compared Zorn and Keros LWD moduli with the Mr values and the secant modu- lus (Ms) based on the permanent strain and resilient strain data obtained from the resilient modulus test. As shown in Figures 65 and 66, strong linear correlations with high R2 values were found between Zorn and Keros LWD moduli and each of the Mr and Ms. White et al. (2007a) compiled ranges of LWD modulus values for various types of cohesive and granular soils under different compaction conditions, which were obtained from FIGURE 65 Relationship between the 200-mm Zorn ELWD and laboratory Mr and Ms (White et al. 2007). FIGURE 66 Relationship between the 200-mm Keros ELWD and laboratory Mr and Ms (White et al. 2007).

66 field testing programs conducted in previous studies. Table 13 presents the mean and COV of the LWD modulus correspond- ing to the range of moisture deviation from optimum and per- cent-relative compaction based on the standard Proctor test. It is clear that the LWD modulus values reported for cohesive soils have higher COV values (ranging from 46% to 71%) compared with those for granular soils (ranging from 5% to 27%). The authors indicated that cohesive soils showed more moisture sensitivity than did granular soils. This was appar- ent from the relatively high COV of LWD moduli for the sandy lean clay soil within a moisture deviation range of 1% and a relative compaction increase of 4%. In a different study, White et al. (2009a) investigated the relationships between the LWD moduli and intelligent com- paction rollers and proof rolling rutting measurements. Two roller-integrated compaction monitoring technologies, namely the compaction meter value (CMV) and the machine drive power (MDP), as well as three types of LWD devices (Dynatest, Zorn, and Keros), were evaluated in the study. Figure 67 pre- sents the relationships between CMV and LWD measurements obtained on granular subgrade and base materials. It is clear that a strong correlation exists between the CMV and LWD deflection and modulus values. However, LWD measurements had better correlation with CMVs when tests were performed in a carefully excavated trench approximately 100 to 150 mm deep (US-10 project). Although the CMV is correlated with a linear regression relationship with LWD modulus values, it also is correlated with a nonlinear power relationship with LWD deflection values. Based on this, the authors suggested that 90% to 120% of the target values criteria used by the Minnesota DOT need to be reviewed for implementing LWD deflection values owing to the nonlinear nature of the relation- ship with CMV. White et al. (2009a) indicated that relation- ships between MDP with LWD measurements obtained from nongranular materials showed positive correlations, although with varying degrees of uncertainty (i.e., R2 values varied from about 0.3 to 0.8). Nonetheless, the relationships between MDP improve when moisture content is included in the regression analysis. The authors found good correlations between the test rolling rut depth and LWD measurements, which are presented in Figure 68. They suggested that the scatter observed in the relationships was partly attributed to soil variability and the differences in influence depth between heavy test rollers (0.6 to 1.2 m) and LWD tests. Virginia Study Hossain and Apeagyei (2010) conducted a study to investi- gate the ability of the LWD to measure in situ modulus of base course materials and subgrade and assess their degree of compaction. The LWD, the GeoGauge, and the DCP were used to test the base course layer and the subgrade in seven pavement sections in five Virginia counties. The authors found that the modulus values measured by the three devices Soil Name USCS Loose Lift Thickness (mm) Moisture Deviation (% Range) Relative Compaction (% Range) LWD Modulus (MPa) (COV) Cohesive Soils Silt ML 300 -2.5 to -3.0 94–98 47 (-) 200 -1.5 to -4.0 96–102 127 (71) Lean clay with sand CL 250 +1.0 to +3.5 87–95 49 (58) 250 -4.0 to + 0.5 86–93 59 (62) Sandy lean clay CL 250 –6.0 to -5.0 84–88 45 (46) 250 -3.0 to -1.5 85–90 65 (58) Cohesionless Soils Well-graded sand with silt SW- SM 360 -5.0 to -3.5 96–99 24 (27) 250 -6.0 to -4.5 96–100 28 (22) Silty gravel with sand GM 350 -0.5 to 0.0 88–90 33 (15) Silty sand with gravel SM 280 -6.0 to -5.5 95–100 33 (8) Poorly graded gravel GP 300 – 95–103 41 (17) Silty sand SM 360 -1.5 to -1.0 91–95 19 (24) Clayey gravel with sand GC 340 -2.0 to -1.5 86–92 37 (12) Well-graded sand with silt SW- SM 200 -0.5 to +2.0 99–102 8 (5) 200 -5.0 to -4.5 99–101 33 (21) 200 -2.5 to -1.5 97–102 27 (33) Source: White et al. (2007a). USCS, Unified Soil Classification System. TABLE 13 RANGE OF LWD MODULUS VALUES PUBLISHED IN LITERATURE

67 FIGURE 67 Correlation between CMV and LWD measurements obtained from field projects with granular materials (White et al. 2009a). FIGURE 68 Relationship between rut depth and LWD measurements from TH-36 project (White et al. 2009a).

68 showed a high spatial variability. In addition, no good cor- relations were found between the LWD moduli and mea- surement of either the GeoGauge or the DCP. Results also showed that the effect of dry density on the measurements of the three devices was not significant. However, the moisture content showed a significant influence on the three in situ test device measurements, especially the LWD. Based on the obtained results, the authors suggested that the LWD not be used in the quality control of construction until further research could be conducted to determine the causes of the high spatial variability and the influence of moisture on the LWD modulus. International Studies Fleming et al. (1988) demonstrated a correlative ratio between the deformation moduli of the German dynamic plate (GDP), LWD, and the FWD of about 0.5. However, Fleming (1998, 2001) reported that his extensive field-stiffness measure- ments on construction sites showed a relatively consistent correlation of 0.6 between the stiffness moduli of the GDP and FWD. In another study, Fleming (2000) conducted field tests to correlate the moduli of three main types of LWD, namely the TFT, GDP, and Prima 100, with that of the FWD. The results showed that although the correlation coefficient between the FWD and Prima 100 moduli (Eq. 40) was close to one, it varied with the other LWD types, as shown in Eq. 41 and Eq. 42. 1.031 (40)Prima100M EFWD = 1.05 to 2.22 (41)M EFWD GDP= 0.76 to 1.32 (42)M EFWD TFT= Fleming et al. (2009) presented a review of LWD use in compaction control of pavement layers and subgrade soil in Europe. They found that the LWD has been increasingly used to test various types of materials before and during construc- tion of major and minor roadways in the United Kingdom. In addition, it was included in the United Kingdom road foun- dation design and construction specifications. The authors indicated that the correlation coefficient between the LWD and FWD moduli was often reported as approximately one, but appeared to be variable and perhaps site dependent. Kamiura et al. (2000) studied the relationship between the LWD and the plate load test measurements for subgrade materials, which contained volcanic soil, silty sand, and mechanically stabilized crushed stone. Based on the results of tests conducted in this study, the authors found the correla- tion in Eq. 43 and indicated that this correlation was affected by grain size of the tested material. k k k LWD LWD)( )(= +Log 0.0031 log 1.12 (43)30 where kLWD = the ratio of stress on loading plate of the LWD to the measured deflection at this stress, and k30 = the ratio of stress on the plate with a diameter of 300 mm for a PLT to the measured deflection at this stress. Pidwerbesky (1997) evaluated the use of FWD and Load- man LWD to predict the performance of unbound granular base course layers and examine the relationship between these devices. A simulated loading and vehicle emulator (SLAVE) was used to load a pavement structure consisting of 90 mm (3.5 in.) of asphalt concrete layer and 200 mm (7.9 in.) of a crushed-rock base course layer on top of a silty clay sub- grade with a CBR of 12% for approximately 1 year. After loading was completed, trenches were cut through the asphalt layer to test the base course and subgrade soil. The authors found that the Loadman LWD was not capable of differen- tiating the moduli of various layers within a multilayered pavement system, but it could provide an indication of the modulus of the tested layer. Based on regression analysis conducted on collected data, a relatively good correlation was obtained between Loadman LWD and FWD moduli, as shown in Figure 69. DYNAMIC CONE PENETROMETER The dynamic cone penetrometer (DCP) was initially devel- oped in South Africa for in situ evaluation of pavement (Kleyn 1975). Since then, the DCP has been used the United States, the United Kingdom, Australia, and New Zealand for site characterization of pavement layers and subgrades. The standard DCP device consists of an upper fixed 575-mm travel rod with an 8-kg falling weight, a lower rod contain- ing an anvil, and a replaceable cone with an apex angle of FIGURE 69 Correlation between Loadman LWD and FWD moduli (Pidwerbesky 1997).

69 60° and a diameter of 20 mm, as shown in Figure 70. The DCP test is conducted according to ASTM D6951 or ASTM D7380, which involves dropping the weight from a 575-mm height and recording the number of blows versus depth. The penetration rate or PR, sometimes referred as the DCP ratio or penetration index (PI), is then calculated. The DCP ratio is defined by the slope of the curve relating to the number of blows to the depth of penetration (in mm/blow) at given linear depth segments. This device costs about $1,500. Repeatability Dai and Kremer (2006) reported DCP test results as a func- tion of test location. At each location, they performed two DCP tests and found that, at some locations, the two test results were very close. Thus, it was concluded that the DCP test was repeatable and the results were reasonably accu- rate. A good repeatability was also reported by Petersen and Peterson (2006). However, Von Quintus et al. (2008) reported a COV of 2.9% to 27.4% for the DCP test results on 10 types of soil at seven different pavement sections. Similarly, Hossain (2010) found relatively higher values of COV (13% to 68%) for DCP measurements. Larsen et al. (2008) indicated that the high COV for DCP measurements was because the readings are influenced by slight variations in moisture and density. Siekmeier et al. (2009) also reported a significant influence of moisture content on the variability of the DCP measurements. Influence Depth The DCP can be used to evaluate compaction of underlying lifts to depths as great as 1.2 m (4 ft) (Mooney et al. 2008). Advantages and Limitations The DCP is simple, durable, economical, and requires mini- mum training and maintenance; it allows for easy access to sites and provides continuous measurements of the in situ strength of pavement sections and the underlying subgrade layers without the need for digging the existing pavement, as is required with other destructive tests (Chen et al. 2001). In addition, the DCP has a standard specification for test- ing, ASTM D6951, and requires no prior calibration. The DCP test is designed to estimate the structural capacity of pavement layers and embankments; it also has the ability to verify both the level and uniformity of compaction, which makes it an excellent tool for quality control of pavement construction. In addition, it can be used to determine the tested layers’ thicknesses to a depth of 1.2 m (4 ft) (Chen et al. 2001). The DCP test has strong correlation with many FIGURE 70 Dynamic cone penetrometer (DCP).

70 strength and stiffness properties of various types of unbound materials, such as the CBR, shear strength, resilient modu- lus, and elastic modulus. Despite its advantages, the DCP has limitations that have been reported in past studies. DCP testing should be lim- ited to materials with a maximum particle size smaller than 51 mm (2 in.) because large aggregate particles may cause the device to tilt, affecting the accuracy of the test results (ASTM D6951). In addition, large particles may cause a sig- nificant increase in the DCP penetration rate that is not repre- sentative of the actual increase in density or strength (Rathje et al. 2006). DCP testing generally requires two people. Cur- rently, the DCP does not have moisture measurement, GPS, or data storage capabilities. Farrag et al. (2005) indicated that the DCP needed to include a drop handle so that during test- ing the upper drop-height-stop does not become loose and slide down, decreasing the drop height. In addition, they rec- ommended a confinement plate be used in granular materi- als to confine the top 2 to 3 in. for better lift measurement. Finally, some studies suggested that the DCP cannot be used in soft clay soils because it may actually advance under its own weight in such soils (Rathje et al. 2006). Synthesis of Past Research Studies Texas Study Chen et al. (1999) conducted DCP and FWD tests on differ- ent subgrade and base materials in more than six districts in Texas. The subgrade resilient modulus was back-calculated from FWD data using EVERCALC. Based on the results, the correlation in Eq. 44 was developed between FWD back- calculated moduli and the DCP penetration rate. In a later study, Chen et al. (2007) developed a new correlation, pre- sented in Eq. 45, based on tests conducted on subgrade and base soils in Texas. 78.05 (44)0.67M DPIFWD = × − 338 for 10 mm blow 60 mm blow (45) 0.39M DPI DPI FWD ( ) ( )= < < − where MFWD = FWD back-calculated moduli (MPa), and DPI = DCP index or DCP penetration rate (mm/blow). Jayawickrama et al. (2000) evaluated DCP use for com- paction control of granular backfill materials for buried struc- tures. Three types of granular backfill materials—concrete gravel, pea gravel, and 50-50 blend (50% concrete gravel and 50% sand)—were compacted using three different com- pactors, namely an impact rammer, vibratory plate, and air tamper. Test results indicated an increase in the DCP blows at greater depth, which was attributed to the effects of the con- fining pressure. However, Jayawickrama et al. (2000) sug- gested that the DCP was capable of differentiating between the compaction equipment and compaction energy levels that were applied to the backfill material. Chen et al. (2001) conducted a study to evaluate the DCP effectiveness in assessing the modulus of compacted sub- grade and base course materials. Sixty DCP and FWD tests were performed on two pavement test sections. Mr laboratory tests were performed on samples obtained during field test- ing. The authors also investigated the effect of mobile load simulator (MLS) loading on the modulus values by conduct- ing field tests before and after loading. Figure 71 presents a comparison between the moduli values obtained based on FWD, DCP, and laboratory test results. The moduli obtained by using DCP test results yielded similar values to those obtained using FWD tests. In addition, the Mr laboratory- measured values were slightly higher than those determined from the DCP and FWD tests. The authors indicated that the average modulus values for the subgrade were approxi- mately the same before and after loading. The moduli for the base layer were reduced after loading, but the reduction was statistically insignificant. Finally, the authors recommended that to achieve 95% confidence level and an error of estimate of less than 20%, a sample size of six DCP tests should be used for routine characterization of base and subgrade layers. Rathje et al. (2006) reported that, in general, the DCP was able to distinguish between locations with smaller and larger dry unit weights for clayey, sandy soils and fine gravel but FIGURE 71 Comparison of moduli from different tests (Chen et al. 2001).

71 did not provide assessments of adequate compaction when compared with direct measurements of dry unit weight. Rathje et al. (2006) indicated that the DCP test could not be performed in coarse gravel aggregate. Minnesota Study The Minnesota DOT was one of the first state DOTs to use the DCP for the evaluation of unbound pavement layers. During the early 1990s, the DCP was used in various proj- ects for locating high-strength layers in pavement structures and identifying weak spots in constructed embankments (Burnham 1993). Burnham (1993) analyzed 700 DCP test results on subgrade, subbase, and base materials to deter- mine limiting DCP penetration rates that corresponded to conditions of “adequate compaction.” In this study, the limiting values were determined to be 7 mm/blow (0.28 in./ blow) for granular materials and 76.2 mm/blow (3 in./ blow) for silty/clayey material. Although moisture content affected the DCP values, the Minnesota DOT specification at the time included limiting penetration rate values with- out any consideration of moisture content. No direct cor- relation between DCP penetration rate and dry unit weight was found in that study. Moreover, it was also determined that DCP penetration rates were not valid over the top few inches of a compacted lift owing to lack of confinement. Results indicated that the DCP test was able to distinguish between the locations with smaller and larger dry unit weights, but the Minnesota DOT criterion for adequate compaction did not agree with the direct measurements of dry unit weight or relative compaction. Oman (2004) collected data from 21 construction proj- ects around Minnesota. A total of 82 locations consisting of different types of unbound granular base materials were tested. Based on the analysis of collected data, a relation- ship between DCP penetration and gradation and moisture content was developed. The collected data also were used to develop trial DCP specifications. The enhanced DCP speci- fications greatly improved the capability of the DCP and reduced testing time. A simple spreadsheet was developed for the specification, which required gradation data, moisture content at the time of testing, and DCP penetration values. Owing to limited testing data, it was concluded that the pro- posed specification be further validated using additional field testing data. Dai and Kremer (2006) attempted to verify and improve the trial Minnesota DOT DCP specifications developed by Oman (2004). Additional field tests were performed and pilot construction projects were implemented. A total of 11 construction projects were selected, and at each project, sev- eral locations were randomly selected for testing. Various devices were used at each location to obtain in situ stiffness, strength, density, and moisture content. Materials under con- sideration included typical granular base materials in Min- nesota as well as reclaimed asphalt material. Data obtained from pilot projects confirmed the previously established rela- tionship between the DCP penetration index, gradation, and moisture content. Moreover, the DCP and sand cone density data collected in four projects showed that the DCP specifi- cations were consistent with the current sand cone density test specifications, which further validated the specifications. The authors indicated that one of the major advantages of the DCP was that it could be applied to materials on which the sand cone density test could not be performed. Petersen and Peterson (2006) conducted DCP tests at 22 locations on the final lift of a 3-ft subcut consisting of a granular borrow compacted using a vibratory compac- tion roller equipped with intelligent compaction technolo- gies. The authors indicated good correlation was obtained between intelligent compaction measurements (i.e., CMV) and the DCP penetration rate for depths between 203 mm (8 in.) and 406 mm (16 in.) when comparison was done on a point-by-point basis. This suggested that DCP measurement depth was close to the influence depth range of the roller sensors. Davich et al. (2006) conducted a study as part of Local Road Research Board Investigation 829 to validate the DCP specification for compaction control of granular materials. DCP and speedy moisture tests were conducted on samples of three types of granular materials, which were compacted inside an open-topped steel cylinder (half of a 55-gallon steel drum) using a procedure similar to that of the standard Proctor test. The results of this study indicated that the DCP specifica- tion should not be limited to three DCP drops because addi- tional drops might be needed to verify the compaction quality for the entire depth of the layer. The authors concluded that the seating requirement was found to be unnecessary for the subbase layer; however, the requirement was still useful for determining the suitability of an aggregate base surface for paving equipment loading. Regarding moisture content during DCP testing, it was concluded that moisture content should be capped at 10%. Three different ranges of moisture contents (less than 5%, between 5% and 7.5%, and between 7.5% and 10%) were recommended during DCP testing. White et al. (2009a) compared DCP results to IC roller CMV and MDP measurements as well as test-rolling rut val- ues. As shown in Figure 72, a fair correlation was obtained between the DCP values and CMVs for granular subgrade and base materials. The DCP penetration rate had a rela- tively weak correlation with the MDP. However, both cor- relations were improved when the moisture content was included in the regression analyses. The authors proposed a method to determine the bearing capacities under the heavy roller wheel using layered bearing capacity analytical solu- tions and DCP profiles. The ultimate bearing capacities determined using this method were empirically related to the measured rut depths at the surface during test rolling. This was used to determine the target DCP penetration rate

72 Colorado Study Mooney et al. (2008) investigated the efficiency of in situ test devices, including the DCP, for quality assurance of Class 1 backfill in mechanically stabilized earth wall and bridge approach embankment. Extensive testing using the DCP was conducted at two construction sites. The results of conducted tests indicated that the DCP penetration index could replace the current density-based compaction method. However, the authors indicated that moisture content should be considered when developing DCP target values used for compaction accep- tance. In addition, the DCP test results need to be corrected when the DCP penetrates through geosynthetic reinforcements placed in mechanically stabilized embankments. Louisiana Studies Abu-Farsakh et al. (2004) evaluated the viability of using the DCP as a tool for stiffness-based QC/QA procedures during and after the construction of pavement layers and embankments. Results showed that the DCP is an excellent and reliable device for evaluating stiffness/strength properties of various types of unbound materials. Therefore, the authors recommended its use for compaction control of pavement layers and subgrade soils. Based on the nonlinear regression analysis that was con- ducted on data collected in this study, strong correlations were found between the DCP penetration rate and the FWD and PLT moduli. Those correlations are provided in Eqs. 48 through 50. M DPI DPI R FWDln 2.04 5.1873ln( ) 3.27 66.67 0.91 (48)2) )( ( )( = + < < = E DPI DPI R PLT i 9770 ( ) 36.9 0.75 3.27 66.67 0.67 (49) 1.6 2) )( ( = − − < < = )( E DPI DPI R PLT R 4374.5 ( ) 14.9 2.16 3.27 66.67 0.72 (50) 2 1.4 2) )( ( = − − < < = )( Regression analysis was also performed to correlate the laboratory CBR and the DCP penetration rate. The following nonlinear regression model was obtained: CBR 2559.44 7.35 1.04 6.31 66.67 0.93 (51) 1.84 2 DPI DPI R ( ) ( ) ( ) = − + + < < = Mohammad et al. (2009) conducted field and laboratory testing programs to develop models that would predict the resilient modulus of subgrade soils from results of the DCP. A total of four soil types (A-4, A-6, A-7-5, and A-7-6) were considered at different moisture-dry density levels. A simple linear regression analysis was first conducted on the combined data set to develop a model that directly of a subgrade soil to avoid rut failures under the test rolling, thus eliminating the need for this test. Mississippi Study George and Uddin (2000) conducted a study for the Mississippi DOT to relate the DCP penetration rate to the resilient modu- lus obtained from laboratory tests and FWD back-calculated moduli for various types of subgrade soils. Manual and auto- matic DCP tests, as well as FWD tests, were performed on fine-grained (A-6) and coarse-grained (A-3 and A-2-6) sub- grade soils at 12 sites in Mississippi. Shelby tube samples were obtained from tested subgrade soils, and resilient modulus laboratory tests were conducted on those samples. The results indicated that measurements from the manual DCP and auto- matic DCP were statistically the same. Two prediction models provided in Eqs. 46 and 47 for fine-grained soil and coarse- grained soils, respectively, were developed. It was concluded that Mr prediction was not only dependent on DPI but also related to the soil’s physical properties, such as dry density and moisture content. 27.86 0.71 (46) 0.144 7.82 1.925 2M DPI LL w Rr dr c ( ) ( )( )= γ +  =− 90.68 log 0.72 (47) 0.305 0.935 0.674 2M DPI c w Rr u dr c( ) ( )=   γ + = − − where Mr = resilient modulus (MPa), DPI = DCP index (mm/blow), gd = dry density, wc = moisture content, LL = liquid limit, PI = plasticity index, and cu = uniformity coefficient. FIGURE 72 Correlations between CMV and in situ point measurements obtained from TH-36 and US-10 field projects with granular soils (White et al. 2009a).

73 predicted the laboratory-measured Mr from the DCP penetration rate. The results of this analysis yielded the model shown in Eq. 52. Figure 73 illustrates the results of the regression analy- sis. A multiple nonlinear regression analysis also was conducted to develop a model that predicted laboratory-measured Mr from the DCP as well as the physical properties of the tested soils. The results of this analysis are presented in Eq. 53 and Figure 74. It is noted that the developed model was able to provide good prediction for the data obtained from a study by George and Uddin (2000) that was not used in the development of the model. M DPI Rr 1045.9 0.90 (52)1.096 2 )()(= = M DPI w R r ) )( ( )( = + + = 3.86 2020.2 1 619.4 1 0.92 (53) 1.46 1.27 2 where Mr = resilient modulus (MPa), DPI = DCP index (mm/blow), and w = water content (%). Florida Study Parker et al. (1998) reported a study in which automated and manual DCP devices were compared. In this study, a series of DCP tests were conducted to evaluate the in situ strength of granular materials and subgrade soils in Florida. No consid- erable difference in the DPI was found when results from the manual and automated DCP were compared. Furthermore, the study indicated that confinement and depth affected the DCP strength measurement of granular materials, whereas the strength measurements of cohesive materials were mini- mally influenced by confinement. 0 0.04 0.08 0.12 1/DCPI1.096 0 20 40 60 80 100 120 140 M ea su re d M r (M Pa ) Up per 95% Pre dic tion Lev el Lo we r 9 5% Pre dic tion Le ve l DCP - Direct model Mr = 1045.9 (1 / DCPI1.096) R2 = 0.9 FIGURE 73 Mr—DCP direct model (Mohammad et al. 2009). 0 20 40 60 80 100 120 140 Measured Resilient Modulus (MPa) 0 20 40 60 80 100 120 140 Pr ed ic te d R es ili en t M od ul us (M Pa ) DCP - Soil Property Model Data Used in Model Development Data Used in Model Verification R2=0.92 FIGURE 74 Mr—DCP soil property model (Mohammad et al. 2009).

74 oped, which was used for compaction control of the remaining 11-m-high embankment. Subsequent construction monitoring and postconstruction evaluation of the bottom ash embank- ment indicated that the developed criterion was very effective. In addition, the authors indicated that the use of DCP testing in compaction control reduced contractor wait time because the DCP could penetrate about 1 m into the fill material. Wisconsin Study Based on tests conducted on natural earthen materials, industrial by-products, chemically stabilized soils, and other materials at 13 construction sites in Wisconsin, Edil and Benson (2005) found that the DCP could be used to assess the compac- tion quality of subgrades by correlating a normalized DCP parameter with relative compaction for the subgrade soil in question. Iowa Studies In a study for the Iowa DOT, Bergeson (1999) investigated the use of the DCP for quality control and acceptance pro- cedures during embankment construction. For cohesive soils, the field data indicated that the stability and shear resistance measured by the DCP increased with an increase in the compaction effort and were reduced as the moisture content increased. However, the DCP results did not cor- relate with moisture content or density measurements. For granular soils, the DCP was found to be an adequate tool for evaluating the in-place density when moisture control was applied to the embankment. Figure 75 shows the variation Indiana Studies Salgado and Yoon (2003) tested various subgrade soils at seven sites in Indiana using the DCP and the NDG. Four sites contained clayey sands, one contained a well-graded sand with clay, and two contained a poorly graded sand. Soil sam- ples were obtained from the sites and tested in the laboratory. The results of this study indicated that despite having a con- siderable scatter in obtained data, a trend appeared to exist between the DCP penetration index and the soils’ physical properties, such that the penetration index decreased as dry density increased and slightly increased as moisture content increased. The authors proposed the model shown in Eq. 54 to predict the dry density for clayey sand soil from the DCP penetration index. They recommended not using the DCP in testing soil with gravel because unrealistic DCP results could be obtained and the penetrometer shaft could be bent. γ = × × ′σ   × γ−10 (54)1.5 0.14 0.5 DPI p d v A W where DPI = DCP index in mm/blow, pA = reference stress (100 kPa), and s′v = effective stress. Siddiki et al. (2008) conducted a study to develop a crite- rion for compaction quality control of a bottom ash embank- ment. DCP tests were conducted during the compaction of a test pad of coal ash. Based on the results, a criterion of 16 blows for every 300-mm-thick layer of bottom ash was devel- FIGURE 75 Relative density compared with DCP index for granular soil (Bergeson 1999).

75 DPI DPI Log CBR 2.465 – 1.12 log or CBR 292 (55)1.12 )(= = Recently, Berney et al. (2013) reported the results of a study conducted at the U.S. Army Corps of Engineers Research and Development Center in Vicksburg, Mississippi, to examine the effectiveness of the DCP as an alternative to the NDG. As shown in Figure 76, a good correlation between the DCP test and NDG dry density measurements was obtained. How- ever, DCP was not recommended because it did not provide a moisture content measurement. International Studies Based on a series of DCP tests conducted on various types of cohesive and granular soils in the United Kingdom, Hunt- ley (1990) suggested a tentative classification system of soil, shown in Tables 14 and 15, based on DCP penetration resistance (blows per 100 mm). However, the author recom- mended the use of classification tables only with consider- able caution until a better understanding of the mechanics of skin friction on the upper drive rods was established. Different correlations were suggested between the DCP penetration rate and the CBR value. Kleyn (1975) conducted DCP tests on 2,000 samples of pavement materials in stan- dard molds directly following CBR determination. Based on Kleyn’s results, the correlation in Eq. 56 was proposed. In a field study, Smith and Pratt (1983) found the correlation pre- sented in Eq. 57. Livneh and Ishai (1987, 1995) conducted a correlative study between DCP values and the in situ CBR values. During this study, both CBR and DCP tests were done of the DCP penetration rate with relative density for granular materials in that study. It was observed that a DCP penetra- tion rate of 35 mm/blow corresponded to a relative density value of 80%, which was selected as the DCP limit for com- paction control of granular materials. However, more study for validation of this limit value was suggested. Gas Technology Institute Study The Gas Technology Institute (GTI) conducted a research project in which several compaction devices, including the DCP, were evaluated (Farrag et al. 2005). Sand, silty clay, and aggregate base were tested using each device. The study found that the DCP provided only general postcom- paction information, such as existence of weak layers or layer boundaries. The DCP penetration rate also did not correlate well with dry unit weight or other compaction parameters. In addition, the DCP provided unreliable data within the top 6 in. of most material tested because of lack of confinement. U.S. Army Corps of Engineers Studies The U.S. Army Corps of Engineers conducted a field DCP study for a wide range of granular and cohesive materials. The results of this study showed a strong correlation, shown in Eq. 55, between the CBR and the DCP penetration ratio (Webster et al. 1992). This equation has been adopted by many state DOTs and appears in the Mechanistic–Empirical Pavement Design Guide (MEPDG) (Webster et al. 1992; Livneh et al. 1995; Siekmeier et al. 2000; Chen et al. 2001). FIGURE 76 Correlation between DCP and dry density obtained using the NDG (Berney et al. 2013).

76 SOIL COMPACTION SUPERVISOR The soil compaction supervisor (SCS), formerly the soil com- paction meter, consists of a disposable sensor that is con- nected by a cable to a battery-powered, handheld control unit, as shown in Figure 77. The SCS works by embedding the sen- sor at the bottom of the soil layer to be compacted. The sen- sor includes piezoelectric transducers that produce a voltage in response to the waves transmitted through the soil from the compaction equipment. The voltage is transferred to the SCS control unit through the connecting cable. The transmitted voltage increases with the increase in the soil stiffness and density owing to compaction. The main function of the SCS is to monitor the voltages from embedded sensors and report when the asymptotic value of stiffness has been reached. A green light on the display indicates that the soil did not reach maximum stiffness value, whereas a red light indicates that the voltage reached its asymptotic value and the compac- tion process should be stopped. The cost of the SCS device is $1,650. Influence Depth Previous studies by the Gas Research Institute indicated that SCS sensors could provide readings to approximately 762 mm (30 in.) of soil thickness above it (Cardenas 2000; Farrag et al. 2005). Advantages and Limitations The SCS device is portable, economical, and can be operated with minimum training. Red and green light signals provide a clear, instant indication of when compaction should be stopped or continued. This signaling system can reduce the risk of over-compaction and save time during compaction by indicating when rolling is no longer needed. However, some limitations of this device have been reported. Farrag et al. (2005) reported that although the red light had good correlation with 90% relative compaction in sand, the cor- relation was weak for clay. Although the SCS box has good durability, the sensors are less durable. The device also does not provide any test results applicable to design or quality in the laboratory on a wide range of undisturbed and com- pacted fine-grained soil samples, with and without satura- tion. Compacted granular soils were tested in flexible molds with variable, controlled lateral pressures. Field tests were conducted on the natural and compacted layers, representing a wide range of potential pavement and subgrade materi- als. The research resulted in the models shown in Eq. 58. Harrison (1989) also suggested Eqs. 59 and 60 to relate CBR to the DCP for different soils. Log CBR 2.62 1.27 log (56)DPI= − Log CBR 2.56 – 1.15 log (57)DPI= Log CBR 2.2 – 0.71 log (58)1.5DPI( )= DPI DPI )( = > Log CBR 2.56 – 1.16 log for clayey-like soil of 10 mm blow (59) Log CBR 2.70 – 1.12 log for granular soil of 10 mm blow (60) DPI DPI ( ) = < Some researchers have attempted to relate DCP results to the elastic modulus of various unbound materials. Table 16 summarizes the main correlations between the DCP and unbound materials modulus values that were reported in international studies. Classification Range of n Values Sand Gravelly sand Very loose <1 <1 <3 Loose 1–2 2–3 3–7 Medium dense 3–7 4–10 8–20 Dense 8–11 11–17 21–33 Very dense >11 >17 >33 Source: Huntley (1990). TABLE 14 SUGGESTED CLASSIFICATION FOR GRANULAR SOIL USING DCP Classification Range of n Values Very soft <1 Soft 1–2 Firm 3–4 Stiff 5–8 Very stiff to hard >8 Source: Huntley (1990). TABLE 15 SUGGESTED CLASSIFICATION FOR COHESIVE SOIL USING DCP

77 tion of each pass of the compaction equipment. Compaction was continued after the SCS red stop signal was displayed, and the density and moisture content were measured when two and four passes were subsequently completed. The results indicated that the average dry density measurements obtained after the red stop signal was displayed increased by less than 2% with additional compactive effort. Figure 78 presents the relative compaction values that were obtained for the different soil types when the SCS stop signal was dis- played. As shown in the figure, all soil types, except low plas- ticity soil, were compacted to at least 95% relative compaction when the SCS displayed the red stop signal. Cardenas (2000) indicated that the low plasticity soil was overly wet when compacted and therefore did not represent optimal compac- tion conditions. In another study, Farrag et al. (2005) investigated the use of the SCS to monitor the compaction of several bell holes and keyholes that were filled and compacted with sand, silty-clay, and stone-base materials. Figures 79 through 81 compare the results of the SCS device with the obtained relative compaction for sand, silty-clay, and stone-based materials, respectively. The results showed that most of the output signals in sand and stone-base soils corresponded to 90% compaction or higher. However, for the silty-clay soil the SCS red stop signal was obtained at relative compac- tion values that were less than 90%. Figures 79 through 81 also show that the SCS device failed to produce signals when soil height was more than 762 mm (30 in.) above the sensor. Juran and Rousset (1999) conducted a field study in which the SCS was used to assess the compaction quality of five test trenches compacted with sandy backfill material. The results of this study indicated that the SCS generally displayed the red stop signal, at relative compaction values that were less than the required 95% value. PORTABLE SEISMIC PROPERTY ANALYZER The PSPA is a portable version of the large, trailer-mounted seismic pavement analyzer (SPA), which was first developed at the University of Texas at El Paso to test both flexible and rigid pavements for early signs of distress and provide general quality control during pavement construction. As control purposes. In addition, the SCS does not have mois- ture measurement, GPS, or good data storage capabilities. Currently, the SCS has no standard procedure. Rathje et al. (2006) also indicated that it has a weak theoretical basis. In addition, the authors noted that the SCS has not been evalu- ated by state DOTs. Thus, there is a lack of evidence of prior success in using it in compaction control of pavement layers and embankments. Synthesis of Past Research Studies This study’s survey results indicated that no research stud- ies have been performed by state DOTs to evaluate the SCS. Most of the studies that evaluated the device were conducted by the Gas Technology Institute. Cardenas (2000) evaluated the efficacy of the SCS in con- trolling the compaction of soil layers. Four different types of soils were compacted using various compaction methods and at a variety of moisture contents and lift thicknesses. Nuclear density gauge measurements were obtained after the comple- Correlation Reference study Log (Es) = 3.05 – 1.07 Log (PR) De Beer (1990) Log (Es) = 3.25 – 0.89 Log (PR) Log (Es) = 3.652 – 1.17 Log (PR) Pen (1990) Log (EPLT) = (-0.88405) Log (PR) + 2.90625 Konard and Lachance (2000) E (in MPa) = 2,224 DCP – 0.99 Chai and Roslie (1998) Log(MFWD) = 3.04785 1.06166*Log(DPI) South Africa (8) DCP is DCP in blows per 300 mm; EPLT is the modulus from plate load test (in MPa); Es is the elastic modulus (in MPa); and PR is the DCP penetration rate (in mm/blow). TABLE 16 SUMMARY OF DCP-MODULUS CORRELATIONS REPORTED IN INTERNATIONAL STUDIES FIGURE 77 Soil compaction supervisor sensor and control unit (MBW 2003).

78 FIGURE 78 Relative compaction at SCS stop signal (Cardenas 2000). FIGURE 79 Comparison between SCS output and relative compaction at various depths in sand (Farrag et al. 2005).

79 The USW method is an offshoot of the spectral analysis of surface waves (SASW) method applied to high-frequency seismic tests. Both methods are based on the measurement of the dispersive nature of the Rayleigh-type surface waves propagating in a layer to determine the shear wave velocity. The main difference between the two methods is that in the USW, the modulus of the tested layer is directly determined without an inversion algorithm. PSPA testing involves activat- ing the source through the laptop to generate high-frequency surface waves that propagate horizontally and are detected and measured by the two receivers. The two receiver outputs are used to compute the Rayleigh wave velocity (VR) at dif- ferent frequencies, which represent the variation of VR with depth. The VR can be used to compute the Young’s modulus using Eq. 61 (Jersey and Edwards 2009). The PSPA software computes an average modulus over the depth measured. It is worth noting that the depth of the material tested by the PSPA can be controlled by adjusting the receivers’ spacing. 2 1.13 – 0.16 (61)2E v VR( )= ρ where r = total mass density, and V = Poisson’s ratio of soil. Repeatability There are limited data on the repeatability of this device. Von Quintus et al. (2008) reported a COV of PSPA modulus rang- ing from 6.0% to 18.5% when testing different pavement materials. Jersey and Edwards (2009) found that COV values shown in Figure 82, the PSPA consists of two receivers (or accelerometers) and a wave source packaged into a hand- portable system. The device is operated by a laptop, which is connected to the hand-carried transducer unit through a cable that carries power to the receivers and wave source and returns the detected signal to the data acquisition board inside it. The PSPA costs from $20,000 to $30,000. Principle of Operation The PSPA principle of operation is based on generating and detecting surface waves in the tested layer and using the ultra- sonic surface wave (USW) method to determine its modulus. FIGURE 81 Comparison between SCS output and relative compaction in stone-base materials (Farrag et al. 2005). FIGURE 80 Comparison between SCS output and relative compaction in silty-clay (Farrag et al. 2005).

80 ture content more than the dry density. The PSPA modulus, in general, increased with increasing dry density of sandy soil. However, there was significant scatter in these data. Finally, for the gravel soils, the authors reported that it was difficult to use these devices for compaction control because there was significant scatter in the PSPA modulus. Nazarian et al. (2006) conducted an implementation proj- ect for the Texas DOT. Based on the results of this study, pro- cedures to measure the seismic moduli of pavement layers in the lab and the field were developed. Protocols for using the PSPA to assess the compaction quality of different pave- ment layers also were presented. The authors found that the seismic modulus correlated well with the resilient modulus design value. Thus, the use of the PSPA allowed for validat- ing the modulus design input value. U.S. Army Corps of Engineers Study Jersey and Edwards (2009) presented the results of a study in which the PSPA, LWD, and GeoGauge tests were conducted on 11 soil test beds that were constructed at the U.S. Army Corps of Engineers Research and Development Center. The results of the different tests conducted in this study are pre- sented in Figure 83. For poorly graded sands (items 1 and 2) the PSPA moduli had similar trends to the GeoGauge and LWD but had approximately twice the magnitude. For fine-grained soils (items 3, 4, and 5), however, there was no observed trend for PSPA moduli. The authors indicated that the PSPA and the other evaluated devices were simple to use and generally obtained repeatable results. However, additional information about the true nature of the modulus measured by these tools was needed to implement their use in compaction control. Joh et al. (2006) tested 21 subgrade soils and 11 base course materials using an accelerated SASW system (similar to the PSPA), the DCP, the PL, and FWD tests. The authors for PSPA measurements ranged between 10% and 21% for sands and 7% and 36% for fine-grained soils. They indicated that PSPA measurements were repeatable for the same loca- tion but varied among test locations, leading to higher COV values. Advantages and Limitations The PSPA is a small, portable, and easy-to-handle device. Testing with the device takes about 15 s to complete. The PSPA allows for monitoring the stiffness changes in the sub- grade and aggregate base at different stages of compaction. In addition, results of lab and field seismic tests are anticipated to be similar for the same materials. This allows for obtain- ing a lab target modulus value that can be used in the field. However, the PSPA device requires a laptop computer in the field for data acquisition and reduction. In addition, a skilled operator is needed to conduct and analyze the data. The PSPA also requires soil-specific calibration, which involves con- ducting complex resonant column-torsional shear laboratory testing. Furthermore, the PSPA-measured modulus does not represent the stress level encountered in the field and may have to be adjusted to account for the design loading fre- quency and strain. The PSPA also does not have a standard test method. Finally, the PSPA is considerably more expen- sive than the NDG and other in situ test devices. Synthesis of Previous Studies Texas Studies Rathje et al. (2006) evaluated the performance of the PSPA as a tool for compaction control of earth embankments and mechanically stabilized earth wall backfill. In the study, the PSPA was used to test five different soils ranging from high plasticity clay to gravel. The results indicated that for clayey soil the PSPA modulus was influenced by the mois- FIGURE 82 PSPA components and data acquisition system (Ellen et al. 2006).

81 A load cell is located above the plate to measure the force applied by the person leaning on the BCD. The BCD works by applying a small repeatable load to a thin plate in con- tact with the compacted material to be tested. Once loaded, the plate bends and the bending strains are instantaneously measured by the strain gauges mounted on the plate. Propri- etary software within the device uses correlations from the field and laboratory to compute the BCD low-strain modulus based on the measured strains. The strain level associated with the BCD measured modulus is on the order of 0.1% (Weidinger and Ge 2009). The BCD has two modes of operation that account for the boundary effects of the Proctor mold that would not occur in the field (Li 2004). A modulus compaction curve in the lab first has to be developed to establish a target modulus from that curve. Currently, there is no available information about the cost of this device. Influence Depth Briaud and Rhee (2009) found that the depth of influence varied with modulus of materials such that it decreased from 121 to 311 mm (3.4 to 12.24 in.) as the modulus increased from 3 to 300 MPa. The authors reported that for materials with a modulus between 5 and 100 MPa, the depth of influ- ence was at least 150 mm (5.9 in.). No other studies have validated these results. Repeatability Briaud and Rhee (2009) evaluated the repeatability of the BCD by testing the same rubber block eight times. The COV of the strain output was found to be 0.5%. Weidinger and Ge (2009) reported a COV of 4% for BCD modulus when testing silt soil samples compacted in the split Proc- tor mold. found that there was a favorable correlation between the DPI and the shear wave velocity measured using the SASW system for both subgrades and base materials. As shown in Figure 84, this correlation became stronger when shear wave velocity increased. However, the authors could not find any strong correlation between SASW modulus and the coefficient of subgrade reaction obtained from the PLT. Finally, the FWD modulus had good correlation with the SASW modulus. BRIAUD COMPACTION DEVICE The BCD consists of a 150-mm-diameter flexible thin plate attached at the bottom of a rod. The plate is instrumented with eight radial and axial strain gauges, as shown in Figure 85. FIGURE 83 Modulus measured by GeoGauge, LWD, and PSPA (Jersey and Edwards 2009). FIGURE 84 DCP index compared with shear wave velocity (Joh et al. 2006).

82 influence depth might affect its efficacy as a tool for compac- tion control (Weidinger and Ge 2009). Synthesis of Past Research Studies NCHRP Highway IDEA Project 118 As part of the NCHRP Highway IDEA program, Briaud and Rhee (2009) reported the results of a project that aimed at improving the design of the previous prototype BCD for compaction control of various unbound materials. The project included conducting PL and BCD tests on 10 test sections with different types of subgrade base materials. The results of those tests are presented in Figure 86. The BCD modulus had an excellent linear correlation with that obtained using the PLT. In addition, to comparing the BCD modulus with the resil- ient modulus, the authors conducted BCD tests on silty clay samples before performing resilient modulus laboratory tests Advantages and Limitations The BCD has several advantages. It is easy to use and can be carried and operated by one person because it weighs only 9.6 kg (4.35 lb). In addition, the BCD is much faster than other in situ test devices, with an actual testing time of approximately 5 s. It also can be used in the lab to determine the target modulus that can be utilized for compaction con- trol of unbound materials in the field. The BCD is a relatively new device, so it has not been extensively evaluated by previous studies, especially those sponsored by state DOTs. One of the main limitations of this device is that it cannot be used for very soft or very stiff soils. For soft soils, the BCD plate simply penetrates into the soil without bending. For stiff soils, the bending of the plate is not adequate for precise measurement of the strains. The device is considered effective in soils with moduli ranging from 5 to 150 MPa (725.2 to 2,175.6). Finally, its relatively shallow FIGURE 85 Photographs and a conceptual sketch of the BCD (Briaud and Rhee 2009).

83 specifications based on both dry density and modulus, which ultimately would result in uniformly dense and strong com- pacted soil layers. However, the authors noted that because of the limitation of the BCD’s influence depth, it would be difficult to effectively assess the soil modulus beyond several inches below the surface. INTELLIGENT COMPACTION All of the aforementioned in situ test devices can assess the mechanical properties of only a very small portion of the compacted materials around the testing location (Kim et al. 2010). Consequently, there may be weak compaction areas unidentified by the limited spot tests. This may result in non- uniform and inadequate compaction, leading to unsatisfactory long-term performance of the compacted layer. To address this issue, research has been performed to assess the quality of compaction along the entire volume of the compacted material using new compaction technologies, such as continuous com- paction control (CCC) and intelligent compaction techniques. The development and evaluation of CCC technologies were initiated in Europe during the late 1970s for use on vibratory rollers compacting granular material (Forssblad 1980; Thurner and Sandström 1980). However, since then CCC technologies have been expanded to different materials and are currently available for different configurations and roller types. The CCC technologies involve using rollers equipped with a real-time kinematic system (RTK), GPS, roller-integrated measurement system, and an onboard, real-time display of all compaction measurements. If the roller has an automatic on them. A total of five samples at five different moisture contents were tested. The authors concluded that there was an excellent correlation between the BCD modulus and the resilient modulus. The authors proposed a procedure for using BCD for com- paction control of soil layers. In this procedure, standard or modified Proctor tests are performed and the optimum mois- ture content and the maximum dry density are determined from the compaction curves. In those tests, the BCD is conducted on top of the Proctor mold sample. This is done to obtain the BCD modulus versus the moisture content curve, which is used to define the maximum BCD modulus and the corresponding optimum moisture content. The authors suggested that the target field modulus value be 75% of the maximum modulus value obtained in the Proctor tests. This target value is verified by conducting BCD tests on the compacted soil in the field. In addition, the moisture content should be verified independently through field testing. Weidinger and Ge (2009) evaluated BCD laboratory pro- cedures for compacted silty soil. That study also compared the BCD modulus to the dynamic and shear moduli determined from ultrasonic pulse velocity tests on the same compacted silt samples. The authors found that the BCD modulus correlated well with the ultrasonic pulse velocity results. In addition, they suggested that conducting BCD tests on the Proctor com- pacted soil was simple and quick, which allowed for develop- ing two important soil trends: the dry density versus moisture content curve and the BCD modulus versus moisture con- tent curve. This could be used to establish field compaction FIGURE 86 Correlation between plate loading test and the BCD test (Briaud and Rhee 2009).

84 it is suggested that all measurements at calibration areas and production areas during quality assurance be obtained at a constant amplitude setting to avoid complication in data analysis and interpretation. Influence Depth The influence depth of the IC roller varies with type of ICMV measurement used. For accelerometer-based measurement systems, the influence depths of measurements were reported to range between 0.8 and 1.5 m (2.62 to 4.92 ft) under a 12-ton vibratory roller (International Society for Soil Mechanics and Geotechnical Engineering 2005; NCHRP 21-09 2009; White et al. 2009). On the other hand, for MDP-based measurements the depth of influence ranged between 0.3 and 0.6 m (1 to 2 ft) depending on the variability of the underlying layer (White et al. 2009a). The ICMV influence depth is affected by roller size, vibration frequency, speed of roller, and the force level that it can generate (Chang et al. 2011). However, Mooney et al. (2011) found that the vibrational amplitude has a minimal effect on the ICMV depths. Advantages and Limitations There are several benefits of IC technologies that have been identified in the literature. IC technologies provide more feedback control for its vibration amplitude and/or frequency, the system is referred to as “intelligent” compaction (IC). During compaction, IC rollers maintain a continuous record of measurements, including the number of roller passes, roller GPS location, IC measurement value (ICMV), and roller vibration amplitudes and/or frequencies. Real-time, onboard, color-coded displays of those measurements provide a spa- tial record of compaction quality and are used to optimize the compaction by adjusting the roller settings manually or auto- matically. Figure 87 presents an example of an IC roller. There currently are seven types of IC single-drum rollers in the United States that are used to compact various types of unbound materials (see www.intelligentcompaction.com). Those rollers use different ICMVs to evaluate the level of compaction. A summary of the ICMV measurements is pro- vided in Table 17. ICMVs are computed based on either the measurements of accelerometers mounted on the roller drum or machine drive power (MDP) measurements. Two different approaches are used to compute ICMVs based on acceler- ometer measurements. The first involves computing the ratio of selected frequency harmonics for a set time interval (e.g., CMV and CCV). The second includes computing stiffness (e.g., ks) or elastic modulus (e.g., Evib) of compacted material based on a drum-ground interaction model and some assump- tions (Chang et al. 2011). ICMVs are influenced by factors such as machine settings (frequency, amplitude). Therefore, FIGURE 87 An example of IC roller, the Bomag VarioControl System (Chang et al. 2011).

85 tests. The compaction information collected using IC roll- ers provides better assessment of the achieved compaction levels because of the significantly larger depth of influence of ICMVs compared with those obtained using in situ tests such as the NDG, LWD, and GeoGauge, as demonstrated in Figure 88. IC technologies also can be especially beneficial for maintaining consistent rolling patterns under lower vis- ibility conditions, such as night paving operations (Chang et al. 2011). efficient and uniform construction process control and QA practice as the rollers map out the stiffness characteristics and quality of compaction for the whole compacted area. Thus, IC provides an effective approach for identifying weak areas in subgrade and base layers that require addi- tional compaction before the placement of surface layers. In addition, over-compaction that can occur during conven- tional compaction can be prevented by using IC because it reduces the number of roller passes and proof-rolling IC Measurements Units IC Systems Model Definition Compaction meter value (CMV) None Caterpillar, Dynapac, Volvo 2ACMV = C A Machine drive power (MDP) None Caterpillar ( ) ' g AMDP = P - Wv sin + - mv + b g Compaction control value (CCV) None Sakai 0.5 1.5 2 2.5 3 0.5 CCV 100A A A A A A A Ω Ω Ω Ω Ω Ω Ω + + + + = × + Stiffness (Kb) MN/m Ammann/Case 2 cos( )o o b d d m eK m z φ ω= + Vibration modulus (Evib) MN/m 2 Bomag 2 1 2 2 2 (2 )2 (1 ) 2.14 0.5ln (1 ) 16 ( ) ( /2) vib vib b e r E aF z a E v v m m m g d pi pi ⋅ ⋅ ⋅∆ = ∆ ⋅ ⋅ ⋅ ⋅ − ⋅ + − ⋅ ⋅ + + ⋅ ⋅ Source: Chang et al. (2011). TABLE 17 SUMMARY OF ICMV MEASUREMENTS FIGURE 88 Illustration of differences in measurement influence depths obtained by using different testing devices (White et al. 2007a).

86 the average MDP measurements and those of different in situ tests. Figure 89 presents the simple linear correlations obtained in one of the projects. Those correlations were improved by incorporating moisture content as a regression variable. The results of this study also showed that the CMV measurements were weakly correlated with machine power for sand. Finally, White et al. (2007a) concluded that a single in situ test point did not provide confidence in representing the average soil engineering property values over a given area. Therefore, multiple tests should be performed to determine soil proper- ties with any degree of confidence. White and Thompson (2007b) evaluated the relationship between the CMV as well as the MDP and various in situ However, there are several disadvantages of the IC sys- tem. IC rollers are more expensive than ordinary ones. IC system measurements are sensitive to moisture; however, currently they are not capable of recording the moisture con- tent of compacted material. There is also no consistent rela- tionship between the ICMVs of different IC systems owing to their different computation algorithms and definitions. This inconsistency in the ICMV definition along with the lack of comprehensive correlations between IC outputs and conven- tional tests are the main obstacles for industry standardiza- tion and the development of IC acceptance specifications. Finally, although IC technologies have been implemented in Europe and Japan for many years, they have been intro- duced to the United States only recently (White et al. 2009a). Therefore, there is still a lack of experience, knowledge, and availability of IC equipment in the United States. These limi- tations may explain the difficulties in implementing IC tech- nologies by state DOTs and paving contractors. Synthesis of Past Research Studies Oklahoma Study Mooney et al. (2003) presented the results of a study that involved monitoring the roller drum vibration during com- paction of well-graded sand test beds. The DCP test was also conducted and used to assess the mechanical properties of the compacted material. The study’s results indicated that the time-domain drum and frame acceleration amplitudes were mildly sensitive to increases in underlying material stiffness properties such that normalized drum acceleration values slightly increased when the DCP penetration index more than doubled. Harmonic content, reported as total harmonic distortion, had more sensitivity to changes in underlying material properties. In addition, it had good correlation with the DCP penetration index, especially when the sublift material was densified and stiffened. Iowa Studies White et al. (2007a, 2009b) conducted several studies to eval- uate the use of CCC technologies in Iowa. White et al. (2006) conducted a study that used CCC rollers at three project sites. For the first two projects, the rollers had internal sensors to monitor the power consumption used to move them. An onboard computer and display screen and a GPS system were used to compact test sections consisting of different types of cohesive soils. In the third project, rollers equipped with CMV and MDP technologies were used to compact clayey and sandy subgrade soils. NDG, GeoGauge, LWD, CH, and DCP tests were conducted during compaction stages in all three projects. The results of this study indicate that the com- paction monitoring technology identified “wet” and “soft” spots incorporated into a test section. In addition, for cohe- sive soil, good correlations were generally obtained between FIGURE 89 MDP correlation with in situ compaction measurements using spatially nearest data pairs (circles) and averaged measurements for given roller pass (squares), kickapoo silt, strip 1 (White et al. 2007a).

87 The authors attributed this to the relative differences in influ- ence depth between CMV measurements and those of the GeoGauge and LWD. However, relatively good correlations were obtained between CMV and DCP measurements for greater depths (200 to 400 mm). White et al. (2007a, 2007b, 2009a) conducted several stud- ies for the Minnesota DOT to evaluate and implement the use of IC in compaction control of pavement layers and sub- grade soils. White et al. (2007a) conducted three field studies at two project sites to investigate the relationships between CCV, MDP, and kB stiffness measurements obtained using three IC rollers and the dry unit weight, soil strength, and modulus parameters determined from NDG, DCP, and LWD, respectively. The authors found strong correlations between kB and in situ test results for test sections with a relatively wide range of material stiffness and comparatively poor cor- relations for sections with more uniform conditions. test device measurements by constructing and testing five strips that consisted of different granular materials. Figure 90 shows the relationships between the average in situ test mea- surements and the CMV and MDP. Although the correlations between CMV and in situ test measurements were linear, the correlations between MDP and these measurements showed a logarithmic relationship. Minnesota Studies Petersen and Peterson (2006) documented an IC demonstra- tion project for the Minnesota DOT and the associated field and laboratory testing. The IC vibratory roller equipped with CMV technology was used to compact 914-mm (3-ft) sub- cut consisting of a select granular borrow material. Results showed that poor correlations between the CMV measure- ments and the GeoGauge and LWD moduli were obtained when the comparison was done on a point-by-point basis. FIGURE 90 Relationships between average in situ and roller integrated compaction measurements (White and Thompson 2007b).

88 when the amplitude was changed. Therefore, measurements obtained at different amplitudes should be treated separately. In another study, White et al. (2009b) found that the CMV was reproducible with variation in nominal speeds between 3.2 and 4.8 km/h. The authors recommended that CMVs be evaluated in conjunction with roller resonant meter values because roller “jumping” affected CMVs. Results of multiple calibration test strips and production areas from one project consisting of clayey subgrade soil showed correlations with varying degrees of uncertainty (i.e., R2 values varied from about 0.3 to 0.8) between the MDP and measurements obtained using the LWD. White et al. (2009a) indicated that MDP values were repeatable with changes in amplitude (from a = 0.85 mm to 1.87 mm) and at a nominal speed of 3.2 km/h. However, MDP values were highly variable when operated at a 6.4 km/h nominal speed. It is worth noting that in another study, White et al. (2009b) indicated that MDP values were affected by the change in amplitude. NCHRP Project 21-09 NCHRP Project 21-09, “Intelligent Soil Compaction Systems,” evaluated several rollers equipped with different types of IC technologies (summary of CCC and IC rollers investigated in NCHRP Project 21-09) and compared their measurements to different in situ and laboratory test results (see Table 18). Simple linear correlations between ICMVs (i.e., MDP, CMV, Evib, ks, CCV) and NDG, LWD, and PLT modulus were pos- sible for a layer underlain by a relatively homogenous and stiff layer. However, different factors were found to affect those correlations: • Heterogeneity in underlying layer support conditions • High moisture content variation • Narrow range of measurements • Machine operation setting variation (e.g., amplitude, frequency, speed) and roller “jumping” • Nonuniform drum/soil contact conditions • Uncertainty in spatial pairing of point measurements and ICMV • Intrinsic measurement errors associated with the ICMV and in situ point measurements. Furthermore, at the project scale average, the dry unit weight, and the DPI values had good correlations with CCV with R2 values of 0.52 and 0.79, respectively. However, poor correlations were found between the LWD moduli and the CCV values. This was attributed to differences in the influence depths of LWD and the IC roller measurements. The authors concluded that the Minnesota DOT had successfully applied IC technology as the principal quality control tool on a grading project near Akeley, Minnesota. The entire project passed the test-rolling acceptance criteria. In another study, White et al. (2009a) documented the results of IC implementation projects. In this study, proof- rolling rut measurements were compared with various IC roller and in situ test measurements at four different sites in Minnesota. The results of measurements obtained in a proj- ect that included granular subgrade materials showed that the LWD modulus and DCP results correlated well with the CMVs when the LWD was performed in a carefully excavated trench 100 to 150 mm (4 to 6 in.) deep, and the DCP first blow was regarded as a seating blow. White et al. (2009a) indicated that the significantly different stress paths for loading under roller and LWD loading were found to be one of the causes of scatter in the relationships between CMV and LWD modulus measurements. As shown in Figure 91, the CMV had a strong correlation with test-rolling rut values. However, poor correla- tion was reported between dry unit weight and CMVs. CMV data obtained from repeated passes demonstrated that the mea- surements were repeatable, but they were not reproducible Rut Depth (mm) FIGURE 91 CMV compared with rut depth (White et al. 2009a). Roller Manufacturer Intelligent Compaction Features Roller-Integrated Measurement Automatic Feedback Control of: GPS-Based Documentation Ammann/Case Stiffness ks Eccentric force, amplitude, and frequency Yes Bomag Stiffness Evb Vertical eccentric for amplitude Yes Caterpillar MDP, CMVC None Yes Dynapac US CMVD Eccentric force amplitude Yes Volvo CMVV None No Sakai America CCV None Yes TABLE 18 SUMMARY OF CCC AND IC ROLLERS INVESTIGATED IN NCHRP PROJECT 21-09

89 on their measurement accuracy. Therefore, several DOTs have attempted to establish those values based on pilot proj- ects or by constructing control strips along a project. Some devices also have limitations on the type of unbound material they can test. In addition, those devices apply different load magnitudes during the test, resulting in different measure- ment results. Although the results of in situ stiffness/strength devices were found to be affected by moisture content, none of these devices have the ability to measure it. The devices possess different influence depth values. Thus, careful con- sideration should be given when analyzing their results and using them for compaction control. Several correlations were developed between the in situ test devices’ measurements and design input parameters, such as the resilient modulus and CBR. However, those correlations are to be used with caution because they can be applied to only certain types of unbound materials and were developed for very specific con- figurations of these devices. In general, no strong correlation was found between in situ stiffness/strength measurements and in-place density because their relationship continuously changes depending on the moisture content. Almost all of the research and implementation conducted by the FHWA and state DOTs focusing on the use of CCC and IC have reported considerable success. However, those agencies are still hesi- tant to widely use these techniques in the field primarily because of the lack of available industry standardization as well as standards and acceptance specifications. Implemen- tation projects in different states can help to address such issues and incorporate CCC and IC into DOT practice. Of all the factors cited, heterogeneity in conditions of under- lying layers was identified as the major factor affecting the correlations. A multiple regression analysis approach was proposed that included ICMV measurements and in situ test results of underlying layers to improve correlations. Mooney et al. (2010) concluded that modulus measure- ments generally captured the variation in ICMVs better than did traditional dry unit weight measurements. In addition, DCP tests were found effective in detecting deeper, “weak” areas that were commonly identified by ICMVs and not by other in situ tests. Finally, it was concluded that relatively constant machine operation settings were critical for calibra- tion strips (e.g., constant amplitude, frequency, and speed), and correlations generally were better for low-amplitude set- tings (e.g., 0.7 to 1.1 mm). SUMMARY The previous sections provide comprehensive details of vari- ous in situ methods that can assess the stiffness or strength of unbound materials and have been used as tools for controlling the quality of their compaction. The CH, DCP, GeoGauge, and LWD have been evaluated by many DOTs. The DCP and LWD are currently the most widely used test devices in field projects for compaction quality control and assurance of base layers, subgrade soils, and embankments. All devices except the PSPA might have difficulties in establishing tar- get field value in the laboratory owing to boundary effects